]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blame - include/linux/skbuff.h
Merge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
[mirror_ubuntu-jammy-kernel.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.
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
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
66 * is TCP or UDP. The IPv4 header may contain IP options
67 * This feature cannot be set in features for a device
68 * with NETIF_F_HW_CSUM also set. This feature is being
69 * DEPRECATED (see below).
70 *
71 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
72 * TCP or UDP packets over IPv6. These are specifically
73 * unencapsulated packets of the form IPv6|TCP or
74 * IPv4|UDP where the Next Header field in the IPv6
75 * header is either TCP or UDP. IPv6 extension headers
76 * are not supported with this feature. This feature
77 * cannot be set in features for a device with
78 * NETIF_F_HW_CSUM also set. This feature is being
79 * DEPRECATED (see below).
80 *
81 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
82 * This flag is used only used to disable the RX checksum
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
88 * verification is in set 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),
118 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
119 * two. If the device were only able to verify the UDP checksum and not
120 * GRE, either because it doesn't support GRE checksum of because GRE
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_
127 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
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
156 * offset of the packet, however they should not attempt to validate that the
157 * checksum refers to a legitimate transport layer checksum-- it is the
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 *
7a6ae71b
TH
181 * This has the same meaning on as CHECKSUM_NONE for checksum offload on
182 * output.
78ea85f1 183 *
7a6ae71b
TH
184 * CHECKSUM_COMPLETE:
185 * Not used in checksum output. If a driver observes a packet with this value
186 * set in skbuff, if should treat as CHECKSUM_NONE being set.
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
dba00306
DC
192 * will set set csum_start and csum_offset accordingly, set ip_summed to
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
203 * accordingly. Note the there is no indication in the skbuff that the
204 * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
205 * both IP checksum offload and FCOE CRC offload must verify which offload
206 * is configured for a packet presumably by inspecting packet headers.
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
214 * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
215 * are set to refer to the outermost checksum being offload (two offloaded
216 * 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
1da177e4 241struct net_device;
716ea3a7 242struct scatterlist;
9c55e01c 243struct pipe_inode_info;
a8f820aa 244struct iov_iter;
fd11a83d 245struct napi_struct;
d58e468b
PP
246struct bpf_prog;
247union bpf_attr;
df5042f4 248struct skb_ext;
1da177e4 249
34666d46 250#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1da177e4 251struct nf_bridge_info {
3eaf4025
FW
252 enum {
253 BRNF_PROTO_UNCHANGED,
254 BRNF_PROTO_8021Q,
255 BRNF_PROTO_PPPOE
7fb48c5b 256 } orig_proto:8;
72b1e5e4
FW
257 u8 pkt_otherhost:1;
258 u8 in_prerouting:1;
259 u8 bridged_dnat:1;
411ffb4f 260 __u16 frag_max_size;
bf1ac5ca 261 struct net_device *physindev;
63cdbc06
FW
262
263 /* always valid & non-NULL from FORWARD on, for physdev match */
264 struct net_device *physoutdev;
7fb48c5b 265 union {
72b1e5e4 266 /* prerouting: detect dnat in orig/reply direction */
72b31f72
BT
267 __be32 ipv4_daddr;
268 struct in6_addr ipv6_daddr;
72b1e5e4
FW
269
270 /* after prerouting + nat detected: store original source
271 * mac since neigh resolution overwrites it, only used while
272 * skb is out in neigh layer.
273 */
274 char neigh_header[8];
72b31f72 275 };
1da177e4
LT
276};
277#endif
278
95a7233c
PB
279#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
280/* Chain in tc_skb_ext will be used to share the tc chain with
281 * ovs recirc_id. It will be set to the current chain by tc
282 * and read by ovs to recirc_id.
283 */
284struct tc_skb_ext {
285 __u32 chain;
286};
287#endif
288
1da177e4
LT
289struct sk_buff_head {
290 /* These two members must be first. */
291 struct sk_buff *next;
292 struct sk_buff *prev;
293
294 __u32 qlen;
295 spinlock_t lock;
296};
297
298struct sk_buff;
299
9d4dde52
IC
300/* To allow 64K frame to be packed as single skb without frag_list we
301 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
302 * buffers which do not start on a page boundary.
303 *
304 * Since GRO uses frags we allocate at least 16 regardless of page
305 * size.
a715dea3 306 */
9d4dde52 307#if (65536/PAGE_SIZE + 1) < 16
eec00954 308#define MAX_SKB_FRAGS 16UL
a715dea3 309#else
9d4dde52 310#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
a715dea3 311#endif
5f74f82e 312extern int sysctl_max_skb_frags;
1da177e4 313
3953c46c
MRL
314/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
315 * segment using its current segmentation instead.
316 */
317#define GSO_BY_FRAGS 0xFFFF
318
8842d285 319typedef struct bio_vec skb_frag_t;
1da177e4 320
161e6137 321/**
7240b60c 322 * skb_frag_size() - Returns the size of a skb fragment
161e6137
PT
323 * @frag: skb fragment
324 */
9e903e08
ED
325static inline unsigned int skb_frag_size(const skb_frag_t *frag)
326{
b8b576a1 327 return frag->bv_len;
9e903e08
ED
328}
329
161e6137 330/**
7240b60c 331 * skb_frag_size_set() - Sets the size of a skb fragment
161e6137
PT
332 * @frag: skb fragment
333 * @size: size of fragment
334 */
9e903e08
ED
335static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
336{
b8b576a1 337 frag->bv_len = size;
9e903e08
ED
338}
339
161e6137 340/**
7240b60c 341 * skb_frag_size_add() - Increments the size of a skb fragment by @delta
161e6137
PT
342 * @frag: skb fragment
343 * @delta: value to add
344 */
9e903e08
ED
345static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
346{
b8b576a1 347 frag->bv_len += delta;
9e903e08
ED
348}
349
161e6137 350/**
7240b60c 351 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
161e6137
PT
352 * @frag: skb fragment
353 * @delta: value to subtract
354 */
9e903e08
ED
355static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
356{
b8b576a1 357 frag->bv_len -= delta;
9e903e08
ED
358}
359
161e6137
PT
360/**
361 * skb_frag_must_loop - Test if %p is a high memory page
362 * @p: fragment's page
363 */
c613c209
WB
364static inline bool skb_frag_must_loop(struct page *p)
365{
366#if defined(CONFIG_HIGHMEM)
367 if (PageHighMem(p))
368 return true;
369#endif
370 return false;
371}
372
373/**
374 * skb_frag_foreach_page - loop over pages in a fragment
375 *
376 * @f: skb frag to operate on
1dfa5bd3 377 * @f_off: offset from start of f->bv_page
c613c209
WB
378 * @f_len: length from f_off to loop over
379 * @p: (temp var) current page
380 * @p_off: (temp var) offset from start of current page,
381 * non-zero only on first page.
382 * @p_len: (temp var) length in current page,
383 * < PAGE_SIZE only on first and last page.
384 * @copied: (temp var) length so far, excluding current p_len.
385 *
386 * A fragment can hold a compound page, in which case per-page
387 * operations, notably kmap_atomic, must be called for each
388 * regular page.
389 */
390#define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
391 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
392 p_off = (f_off) & (PAGE_SIZE - 1), \
393 p_len = skb_frag_must_loop(p) ? \
394 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
395 copied = 0; \
396 copied < f_len; \
397 copied += p_len, p++, p_off = 0, \
398 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
399
ac45f602
PO
400#define HAVE_HW_TIME_STAMP
401
402/**
d3a21be8 403 * struct skb_shared_hwtstamps - hardware time stamps
ac45f602
PO
404 * @hwtstamp: hardware time stamp transformed into duration
405 * since arbitrary point in time
ac45f602
PO
406 *
407 * Software time stamps generated by ktime_get_real() are stored in
4d276eb6 408 * skb->tstamp.
ac45f602
PO
409 *
410 * hwtstamps can only be compared against other hwtstamps from
411 * the same device.
412 *
413 * This structure is attached to packets as part of the
414 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
415 */
416struct skb_shared_hwtstamps {
417 ktime_t hwtstamp;
ac45f602
PO
418};
419
2244d07b
OH
420/* Definitions for tx_flags in struct skb_shared_info */
421enum {
422 /* generate hardware time stamp */
423 SKBTX_HW_TSTAMP = 1 << 0,
424
e7fd2885 425 /* generate software time stamp when queueing packet to NIC */
2244d07b
OH
426 SKBTX_SW_TSTAMP = 1 << 1,
427
428 /* device driver is going to provide hardware time stamp */
429 SKBTX_IN_PROGRESS = 1 << 2,
430
a6686f2f 431 /* device driver supports TX zero-copy buffers */
62b1a8ab 432 SKBTX_DEV_ZEROCOPY = 1 << 3,
6e3e939f
JB
433
434 /* generate wifi status information (where possible) */
62b1a8ab 435 SKBTX_WIFI_STATUS = 1 << 4,
c9af6db4
PS
436
437 /* This indicates at least one fragment might be overwritten
438 * (as in vmsplice(), sendfile() ...)
439 * If we need to compute a TX checksum, we'll need to copy
440 * all frags to avoid possible bad checksum
441 */
442 SKBTX_SHARED_FRAG = 1 << 5,
e7fd2885
WB
443
444 /* generate software time stamp when entering packet scheduling */
445 SKBTX_SCHED_TSTAMP = 1 << 6,
a6686f2f
SM
446};
447
52267790 448#define SKBTX_ZEROCOPY_FRAG (SKBTX_DEV_ZEROCOPY | SKBTX_SHARED_FRAG)
e1c8a607 449#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
0a2cf20c 450 SKBTX_SCHED_TSTAMP)
f24b9be5
WB
451#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
452
a6686f2f
SM
453/*
454 * The callback notifies userspace to release buffers when skb DMA is done in
455 * lower device, the skb last reference should be 0 when calling this.
e19d6763
MT
456 * The zerocopy_success argument is true if zero copy transmit occurred,
457 * false on data copy or out of memory error caused by data copy attempt.
ca8f4fb2
MT
458 * The ctx field is used to track device context.
459 * The desc field is used to track userspace buffer index.
a6686f2f
SM
460 */
461struct ubuf_info {
e19d6763 462 void (*callback)(struct ubuf_info *, bool zerocopy_success);
4ab6c99d
WB
463 union {
464 struct {
465 unsigned long desc;
466 void *ctx;
467 };
468 struct {
469 u32 id;
470 u16 len;
471 u16 zerocopy:1;
472 u32 bytelen;
473 };
474 };
c1d1b437 475 refcount_t refcnt;
a91dbff5
WB
476
477 struct mmpin {
478 struct user_struct *user;
479 unsigned int num_pg;
480 } mmp;
ac45f602
PO
481};
482
52267790
WB
483#define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
484
6f89dbce
SV
485int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
486void mm_unaccount_pinned_pages(struct mmpin *mmp);
487
52267790 488struct ubuf_info *sock_zerocopy_alloc(struct sock *sk, size_t size);
4ab6c99d
WB
489struct ubuf_info *sock_zerocopy_realloc(struct sock *sk, size_t size,
490 struct ubuf_info *uarg);
52267790
WB
491
492static inline void sock_zerocopy_get(struct ubuf_info *uarg)
493{
c1d1b437 494 refcount_inc(&uarg->refcnt);
52267790
WB
495}
496
497void sock_zerocopy_put(struct ubuf_info *uarg);
52900d22 498void sock_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
52267790
WB
499
500void sock_zerocopy_callback(struct ubuf_info *uarg, bool success);
501
b5947e5d 502int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len);
52267790
WB
503int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
504 struct msghdr *msg, int len,
505 struct ubuf_info *uarg);
506
1da177e4
LT
507/* This data is invariant across clones and lives at
508 * the end of the header data, ie. at skb->end.
509 */
510struct skb_shared_info {
de8f3a83
DB
511 __u8 __unused;
512 __u8 meta_len;
513 __u8 nr_frags;
9f42f126 514 __u8 tx_flags;
7967168c
HX
515 unsigned short gso_size;
516 /* Warning: this field is not always filled in (UFO)! */
517 unsigned short gso_segs;
1da177e4 518 struct sk_buff *frag_list;
ac45f602 519 struct skb_shared_hwtstamps hwtstamps;
7f564528 520 unsigned int gso_type;
09c2d251 521 u32 tskey;
ec7d2f2c
ED
522
523 /*
524 * Warning : all fields before dataref are cleared in __alloc_skb()
525 */
526 atomic_t dataref;
527
69e3c75f
JB
528 /* Intermediate layers must ensure that destructor_arg
529 * remains valid until skb destructor */
530 void * destructor_arg;
a6686f2f 531
fed66381
ED
532 /* must be last field, see pskb_expand_head() */
533 skb_frag_t frags[MAX_SKB_FRAGS];
1da177e4
LT
534};
535
536/* We divide dataref into two halves. The higher 16 bits hold references
537 * to the payload part of skb->data. The lower 16 bits hold references to
334a8132
PM
538 * the entire skb->data. A clone of a headerless skb holds the length of
539 * the header in skb->hdr_len.
1da177e4
LT
540 *
541 * All users must obey the rule that the skb->data reference count must be
542 * greater than or equal to the payload reference count.
543 *
544 * Holding a reference to the payload part means that the user does not
545 * care about modifications to the header part of skb->data.
546 */
547#define SKB_DATAREF_SHIFT 16
548#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
549
d179cd12
DM
550
551enum {
c8753d55
VS
552 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
553 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
554 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
d179cd12
DM
555};
556
7967168c
HX
557enum {
558 SKB_GSO_TCPV4 = 1 << 0,
576a30eb
HX
559
560 /* This indicates the skb is from an untrusted source. */
d9d30adf 561 SKB_GSO_DODGY = 1 << 1,
b0da8537
MC
562
563 /* This indicates the tcp segment has CWR set. */
d9d30adf 564 SKB_GSO_TCP_ECN = 1 << 2,
f83ef8c0 565
d9d30adf 566 SKB_GSO_TCP_FIXEDID = 1 << 3,
01d5b2fc 567
d9d30adf 568 SKB_GSO_TCPV6 = 1 << 4,
68c33163 569
d9d30adf 570 SKB_GSO_FCOE = 1 << 5,
73136267 571
d9d30adf 572 SKB_GSO_GRE = 1 << 6,
0d89d203 573
d9d30adf 574 SKB_GSO_GRE_CSUM = 1 << 7,
cb32f511 575
d9d30adf 576 SKB_GSO_IPXIP4 = 1 << 8,
61c1db7f 577
d9d30adf 578 SKB_GSO_IPXIP6 = 1 << 9,
0f4f4ffa 579
d9d30adf 580 SKB_GSO_UDP_TUNNEL = 1 << 10,
4749c09c 581
d9d30adf 582 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
cbc53e08 583
d9d30adf 584 SKB_GSO_PARTIAL = 1 << 12,
802ab55a 585
d9d30adf 586 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
90017acc 587
d9d30adf 588 SKB_GSO_SCTP = 1 << 14,
c7ef8f0c 589
d9d30adf 590 SKB_GSO_ESP = 1 << 15,
0c19f846
WB
591
592 SKB_GSO_UDP = 1 << 16,
ee80d1eb
WB
593
594 SKB_GSO_UDP_L4 = 1 << 17,
7967168c
HX
595};
596
2e07fa9c
ACM
597#if BITS_PER_LONG > 32
598#define NET_SKBUFF_DATA_USES_OFFSET 1
599#endif
600
601#ifdef NET_SKBUFF_DATA_USES_OFFSET
602typedef unsigned int sk_buff_data_t;
603#else
604typedef unsigned char *sk_buff_data_t;
605#endif
606
161e6137 607/**
1da177e4
LT
608 * struct sk_buff - socket buffer
609 * @next: Next buffer in list
610 * @prev: Previous buffer in list
363ec392 611 * @tstamp: Time we arrived/left
56b17425 612 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
d84e0bd7 613 * @sk: Socket we are owned by
1da177e4 614 * @dev: Device we arrived on/are leaving by
d84e0bd7 615 * @cb: Control buffer. Free for use by every layer. Put private vars here
7fee226a 616 * @_skb_refdst: destination entry (with norefcount bit)
67be2dd1 617 * @sp: the security path, used for xfrm
1da177e4
LT
618 * @len: Length of actual data
619 * @data_len: Data length
620 * @mac_len: Length of link layer header
334a8132 621 * @hdr_len: writable header length of cloned skb
663ead3b
HX
622 * @csum: Checksum (must include start/offset pair)
623 * @csum_start: Offset from skb->head where checksumming should start
624 * @csum_offset: Offset from csum_start where checksum should be stored
d84e0bd7 625 * @priority: Packet queueing priority
60ff7467 626 * @ignore_df: allow local fragmentation
1da177e4 627 * @cloned: Head may be cloned (check refcnt to be sure)
d84e0bd7 628 * @ip_summed: Driver fed us an IP checksum
1da177e4
LT
629 * @nohdr: Payload reference only, must not modify header
630 * @pkt_type: Packet class
c83c2486 631 * @fclone: skbuff clone status
c83c2486 632 * @ipvs_property: skbuff is owned by ipvs
875e8939
IS
633 * @offload_fwd_mark: Packet was L2-forwarded in hardware
634 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
e7246e12 635 * @tc_skip_classify: do not classify packet. set by IFB device
8dc07fdb 636 * @tc_at_ingress: used within tc_classify to distinguish in/egress
bc31c905
WB
637 * @tc_redirected: packet was redirected by a tc action
638 * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
31729363
RD
639 * @peeked: this packet has been seen already, so stats have been
640 * done for it, don't do them again
ba9dda3a 641 * @nf_trace: netfilter packet trace flag
d84e0bd7
DB
642 * @protocol: Packet protocol from driver
643 * @destructor: Destruct function
e2080072 644 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
a9e419dc 645 * @_nfct: Associated connection, if any (with nfctinfo bits)
1da177e4 646 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
8964be4a 647 * @skb_iif: ifindex of device we arrived on
1da177e4 648 * @tc_index: Traffic control index
61b905da 649 * @hash: the packet hash
d84e0bd7 650 * @queue_mapping: Queue mapping for multiqueue devices
8b700862 651 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
df5042f4 652 * @active_extensions: active extensions (skb_ext_id types)
553a5672 653 * @ndisc_nodetype: router type (from link layer)
d84e0bd7 654 * @ooo_okay: allow the mapping of a socket to a queue to be changed
61b905da 655 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
4ca2462e 656 * ports.
a3b18ddb 657 * @sw_hash: indicates hash was computed in software stack
6e3e939f
JB
658 * @wifi_acked_valid: wifi_acked was set
659 * @wifi_acked: whether frame was acked on wifi or not
3bdc0eba 660 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
dba00306 661 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
4ff06203 662 * @dst_pending_confirm: need to confirm neighbour
a48d189e 663 * @decrypted: Decrypted SKB
161e6137 664 * @napi_id: id of the NAPI struct this skb came from
984bc16c 665 * @secmark: security marking
d84e0bd7 666 * @mark: Generic packet mark
86a9bad3 667 * @vlan_proto: vlan encapsulation protocol
6aa895b0 668 * @vlan_tci: vlan tag control information
0d89d203 669 * @inner_protocol: Protocol (encapsulation)
6a674e9c
JG
670 * @inner_transport_header: Inner transport layer header (encapsulation)
671 * @inner_network_header: Network layer header (encapsulation)
aefbd2b3 672 * @inner_mac_header: Link layer header (encapsulation)
d84e0bd7
DB
673 * @transport_header: Transport layer header
674 * @network_header: Network layer header
675 * @mac_header: Link layer header
676 * @tail: Tail pointer
677 * @end: End pointer
678 * @head: Head of buffer
679 * @data: Data head pointer
680 * @truesize: Buffer size
681 * @users: User count - see {datagram,tcp}.c
df5042f4 682 * @extensions: allocated extensions, valid if active_extensions is nonzero
1da177e4
LT
683 */
684
685struct sk_buff {
363ec392 686 union {
56b17425
ED
687 struct {
688 /* These two members must be first. */
689 struct sk_buff *next;
690 struct sk_buff *prev;
691
692 union {
bffa72cf
ED
693 struct net_device *dev;
694 /* Some protocols might use this space to store information,
695 * while device pointer would be NULL.
696 * UDP receive path is one user.
697 */
698 unsigned long dev_scratch;
56b17425
ED
699 };
700 };
fa0f5273 701 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
d4546c25 702 struct list_head list;
363ec392 703 };
fa0f5273
PO
704
705 union {
706 struct sock *sk;
707 int ip_defrag_offset;
708 };
1da177e4 709
c84d9490 710 union {
bffa72cf 711 ktime_t tstamp;
d3edd06e 712 u64 skb_mstamp_ns; /* earliest departure time */
c84d9490 713 };
1da177e4
LT
714 /*
715 * This is the control buffer. It is free to use for every
716 * layer. Please put your private variables there. If you
717 * want to keep them across layers you have to do a skb_clone()
718 * first. This is owned by whoever has the skb queued ATM.
719 */
da3f5cf1 720 char cb[48] __aligned(8);
1da177e4 721
e2080072
ED
722 union {
723 struct {
724 unsigned long _skb_refdst;
725 void (*destructor)(struct sk_buff *skb);
726 };
727 struct list_head tcp_tsorted_anchor;
728 };
729
b1937227 730#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 731 unsigned long _nfct;
da3f5cf1 732#endif
1da177e4 733 unsigned int len,
334a8132
PM
734 data_len;
735 __u16 mac_len,
736 hdr_len;
b1937227
ED
737
738 /* Following fields are _not_ copied in __copy_skb_header()
739 * Note that queue_mapping is here mostly to fill a hole.
740 */
b1937227 741 __u16 queue_mapping;
36bbef52
DB
742
743/* if you move cloned around you also must adapt those constants */
744#ifdef __BIG_ENDIAN_BITFIELD
745#define CLONED_MASK (1 << 7)
746#else
747#define CLONED_MASK 1
748#endif
749#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
750
751 __u8 __cloned_offset[0];
b1937227 752 __u8 cloned:1,
6869c4d8 753 nohdr:1,
b84f4cc9 754 fclone:2,
a59322be 755 peeked:1,
b1937227 756 head_frag:1,
8b700862 757 pfmemalloc:1;
df5042f4
FW
758#ifdef CONFIG_SKB_EXTENSIONS
759 __u8 active_extensions;
760#endif
b1937227
ED
761 /* fields enclosed in headers_start/headers_end are copied
762 * using a single memcpy() in __copy_skb_header()
763 */
ebcf34f3 764 /* private: */
b1937227 765 __u32 headers_start[0];
ebcf34f3 766 /* public: */
4031ae6e 767
233577a2
HFS
768/* if you move pkt_type around you also must adapt those constants */
769#ifdef __BIG_ENDIAN_BITFIELD
770#define PKT_TYPE_MAX (7 << 5)
771#else
772#define PKT_TYPE_MAX 7
1da177e4 773#endif
233577a2 774#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
fe55f6d5 775
233577a2 776 __u8 __pkt_type_offset[0];
b1937227 777 __u8 pkt_type:3;
b1937227 778 __u8 ignore_df:1;
b1937227
ED
779 __u8 nf_trace:1;
780 __u8 ip_summed:2;
3853b584 781 __u8 ooo_okay:1;
8b700862 782
61b905da 783 __u8 l4_hash:1;
a3b18ddb 784 __u8 sw_hash:1;
6e3e939f
JB
785 __u8 wifi_acked_valid:1;
786 __u8 wifi_acked:1;
3bdc0eba 787 __u8 no_fcs:1;
77cffe23 788 /* Indicates the inner headers are valid in the skbuff. */
6a674e9c 789 __u8 encapsulation:1;
7e2b10c1 790 __u8 encap_hdr_csum:1;
5d0c2b95 791 __u8 csum_valid:1;
8b700862 792
0c4b2d37
MM
793#ifdef __BIG_ENDIAN_BITFIELD
794#define PKT_VLAN_PRESENT_BIT 7
795#else
796#define PKT_VLAN_PRESENT_BIT 0
797#endif
798#define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset)
799 __u8 __pkt_vlan_present_offset[0];
800 __u8 vlan_present:1;
7e3cead5 801 __u8 csum_complete_sw:1;
b1937227 802 __u8 csum_level:2;
dba00306 803 __u8 csum_not_inet:1;
4ff06203 804 __u8 dst_pending_confirm:1;
b1937227
ED
805#ifdef CONFIG_IPV6_NDISC_NODETYPE
806 __u8 ndisc_nodetype:2;
807#endif
8b700862 808
0c4b2d37 809 __u8 ipvs_property:1;
8bce6d7d 810 __u8 inner_protocol_type:1;
e585f236 811 __u8 remcsum_offload:1;
6bc506b4
IS
812#ifdef CONFIG_NET_SWITCHDEV
813 __u8 offload_fwd_mark:1;
875e8939 814 __u8 offload_l3_fwd_mark:1;
6bc506b4 815#endif
e7246e12
WB
816#ifdef CONFIG_NET_CLS_ACT
817 __u8 tc_skip_classify:1;
8dc07fdb 818 __u8 tc_at_ingress:1;
bc31c905
WB
819 __u8 tc_redirected:1;
820 __u8 tc_from_ingress:1;
e7246e12 821#endif
a48d189e
SB
822#ifdef CONFIG_TLS_DEVICE
823 __u8 decrypted:1;
824#endif
b1937227
ED
825
826#ifdef CONFIG_NET_SCHED
827 __u16 tc_index; /* traffic control index */
b1937227 828#endif
fe55f6d5 829
b1937227
ED
830 union {
831 __wsum csum;
832 struct {
833 __u16 csum_start;
834 __u16 csum_offset;
835 };
836 };
837 __u32 priority;
838 int skb_iif;
839 __u32 hash;
840 __be16 vlan_proto;
841 __u16 vlan_tci;
2bd82484
ED
842#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
843 union {
844 unsigned int napi_id;
845 unsigned int sender_cpu;
846 };
97fc2f08 847#endif
984bc16c 848#ifdef CONFIG_NETWORK_SECMARK
6bc506b4 849 __u32 secmark;
0c4f691f 850#endif
0c4f691f 851
3b885787
NH
852 union {
853 __u32 mark;
16fad69c 854 __u32 reserved_tailroom;
3b885787 855 };
1da177e4 856
8bce6d7d
TH
857 union {
858 __be16 inner_protocol;
859 __u8 inner_ipproto;
860 };
861
1a37e412
SH
862 __u16 inner_transport_header;
863 __u16 inner_network_header;
864 __u16 inner_mac_header;
b1937227
ED
865
866 __be16 protocol;
1a37e412
SH
867 __u16 transport_header;
868 __u16 network_header;
869 __u16 mac_header;
b1937227 870
ebcf34f3 871 /* private: */
b1937227 872 __u32 headers_end[0];
ebcf34f3 873 /* public: */
b1937227 874
1da177e4 875 /* These elements must be at the end, see alloc_skb() for details. */
27a884dc 876 sk_buff_data_t tail;
4305b541 877 sk_buff_data_t end;
1da177e4 878 unsigned char *head,
4305b541 879 *data;
27a884dc 880 unsigned int truesize;
63354797 881 refcount_t users;
df5042f4
FW
882
883#ifdef CONFIG_SKB_EXTENSIONS
884 /* only useable after checking ->active_extensions != 0 */
885 struct skb_ext *extensions;
886#endif
1da177e4
LT
887};
888
889#ifdef __KERNEL__
890/*
891 * Handling routines are only of interest to the kernel
892 */
1da177e4 893
c93bdd0e
MG
894#define SKB_ALLOC_FCLONE 0x01
895#define SKB_ALLOC_RX 0x02
fd11a83d 896#define SKB_ALLOC_NAPI 0x04
c93bdd0e 897
161e6137
PT
898/**
899 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
900 * @skb: buffer
901 */
c93bdd0e
MG
902static inline bool skb_pfmemalloc(const struct sk_buff *skb)
903{
904 return unlikely(skb->pfmemalloc);
905}
906
7fee226a
ED
907/*
908 * skb might have a dst pointer attached, refcounted or not.
909 * _skb_refdst low order bit is set if refcount was _not_ taken
910 */
911#define SKB_DST_NOREF 1UL
912#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
913
914/**
915 * skb_dst - returns skb dst_entry
916 * @skb: buffer
917 *
918 * Returns skb dst_entry, regardless of reference taken or not.
919 */
adf30907
ED
920static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
921{
161e6137 922 /* If refdst was not refcounted, check we still are in a
7fee226a
ED
923 * rcu_read_lock section
924 */
925 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
926 !rcu_read_lock_held() &&
927 !rcu_read_lock_bh_held());
928 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
adf30907
ED
929}
930
7fee226a
ED
931/**
932 * skb_dst_set - sets skb dst
933 * @skb: buffer
934 * @dst: dst entry
935 *
936 * Sets skb dst, assuming a reference was taken on dst and should
937 * be released by skb_dst_drop()
938 */
adf30907
ED
939static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
940{
7fee226a
ED
941 skb->_skb_refdst = (unsigned long)dst;
942}
943
932bc4d7
JA
944/**
945 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
946 * @skb: buffer
947 * @dst: dst entry
948 *
949 * Sets skb dst, assuming a reference was not taken on dst.
950 * If dst entry is cached, we do not take reference and dst_release
951 * will be avoided by refdst_drop. If dst entry is not cached, we take
952 * reference, so that last dst_release can destroy the dst immediately.
953 */
954static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
955{
dbfc4fb7
HFS
956 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
957 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
932bc4d7 958}
7fee226a
ED
959
960/**
25985edc 961 * skb_dst_is_noref - Test if skb dst isn't refcounted
7fee226a
ED
962 * @skb: buffer
963 */
964static inline bool skb_dst_is_noref(const struct sk_buff *skb)
965{
966 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
adf30907
ED
967}
968
161e6137
PT
969/**
970 * skb_rtable - Returns the skb &rtable
971 * @skb: buffer
972 */
511c3f92
ED
973static inline struct rtable *skb_rtable(const struct sk_buff *skb)
974{
adf30907 975 return (struct rtable *)skb_dst(skb);
511c3f92
ED
976}
977
8b10cab6
JHS
978/* For mangling skb->pkt_type from user space side from applications
979 * such as nft, tc, etc, we only allow a conservative subset of
980 * possible pkt_types to be set.
981*/
982static inline bool skb_pkt_type_ok(u32 ptype)
983{
984 return ptype <= PACKET_OTHERHOST;
985}
986
161e6137
PT
987/**
988 * skb_napi_id - Returns the skb's NAPI id
989 * @skb: buffer
990 */
90b602f8
ML
991static inline unsigned int skb_napi_id(const struct sk_buff *skb)
992{
993#ifdef CONFIG_NET_RX_BUSY_POLL
994 return skb->napi_id;
995#else
996 return 0;
997#endif
998}
999
161e6137
PT
1000/**
1001 * skb_unref - decrement the skb's reference count
1002 * @skb: buffer
1003 *
1004 * Returns true if we can free the skb.
1005 */
3889a803
PA
1006static inline bool skb_unref(struct sk_buff *skb)
1007{
1008 if (unlikely(!skb))
1009 return false;
63354797 1010 if (likely(refcount_read(&skb->users) == 1))
3889a803 1011 smp_rmb();
63354797 1012 else if (likely(!refcount_dec_and_test(&skb->users)))
3889a803
PA
1013 return false;
1014
1015 return true;
1016}
1017
0a463c78 1018void skb_release_head_state(struct sk_buff *skb);
7965bd4d
JP
1019void kfree_skb(struct sk_buff *skb);
1020void kfree_skb_list(struct sk_buff *segs);
6413139d 1021void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
7965bd4d
JP
1022void skb_tx_error(struct sk_buff *skb);
1023void consume_skb(struct sk_buff *skb);
ca2c1418 1024void __consume_stateless_skb(struct sk_buff *skb);
7965bd4d 1025void __kfree_skb(struct sk_buff *skb);
d7e8883c 1026extern struct kmem_cache *skbuff_head_cache;
bad43ca8 1027
7965bd4d
JP
1028void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1029bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1030 bool *fragstolen, int *delta_truesize);
bad43ca8 1031
7965bd4d
JP
1032struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1033 int node);
2ea2f62c 1034struct sk_buff *__build_skb(void *data, unsigned int frag_size);
7965bd4d 1035struct sk_buff *build_skb(void *data, unsigned int frag_size);
ba0509b6
JDB
1036struct sk_buff *build_skb_around(struct sk_buff *skb,
1037 void *data, unsigned int frag_size);
161e6137
PT
1038
1039/**
1040 * alloc_skb - allocate a network buffer
1041 * @size: size to allocate
1042 * @priority: allocation mask
1043 *
1044 * This function is a convenient wrapper around __alloc_skb().
1045 */
d179cd12 1046static inline struct sk_buff *alloc_skb(unsigned int size,
dd0fc66f 1047 gfp_t priority)
d179cd12 1048{
564824b0 1049 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
d179cd12
DM
1050}
1051
2e4e4410
ED
1052struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1053 unsigned long data_len,
1054 int max_page_order,
1055 int *errcode,
1056 gfp_t gfp_mask);
da29e4b4 1057struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
2e4e4410 1058
d0bf4a9e
ED
1059/* Layout of fast clones : [skb1][skb2][fclone_ref] */
1060struct sk_buff_fclones {
1061 struct sk_buff skb1;
1062
1063 struct sk_buff skb2;
1064
2638595a 1065 refcount_t fclone_ref;
d0bf4a9e
ED
1066};
1067
1068/**
1069 * skb_fclone_busy - check if fclone is busy
293de7de 1070 * @sk: socket
d0bf4a9e
ED
1071 * @skb: buffer
1072 *
bda13fed 1073 * Returns true if skb is a fast clone, and its clone is not freed.
39bb5e62
ED
1074 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1075 * so we also check that this didnt happen.
d0bf4a9e 1076 */
39bb5e62
ED
1077static inline bool skb_fclone_busy(const struct sock *sk,
1078 const struct sk_buff *skb)
d0bf4a9e
ED
1079{
1080 const struct sk_buff_fclones *fclones;
1081
1082 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1083
1084 return skb->fclone == SKB_FCLONE_ORIG &&
2638595a 1085 refcount_read(&fclones->fclone_ref) > 1 &&
39bb5e62 1086 fclones->skb2.sk == sk;
d0bf4a9e
ED
1087}
1088
161e6137
PT
1089/**
1090 * alloc_skb_fclone - allocate a network buffer from fclone cache
1091 * @size: size to allocate
1092 * @priority: allocation mask
1093 *
1094 * This function is a convenient wrapper around __alloc_skb().
1095 */
d179cd12 1096static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
dd0fc66f 1097 gfp_t priority)
d179cd12 1098{
c93bdd0e 1099 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
d179cd12
DM
1100}
1101
7965bd4d 1102struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
b0768a86 1103void skb_headers_offset_update(struct sk_buff *skb, int off);
7965bd4d
JP
1104int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1105struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
08303c18 1106void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
7965bd4d 1107struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
bad93e9d
OP
1108struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1109 gfp_t gfp_mask, bool fclone);
1110static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1111 gfp_t gfp_mask)
1112{
1113 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1114}
7965bd4d
JP
1115
1116int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1117struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1118 unsigned int headroom);
1119struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1120 int newtailroom, gfp_t priority);
48a1df65
JD
1121int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1122 int offset, int len);
1123int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1124 int offset, int len);
7965bd4d 1125int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
cd0a137a
FF
1126int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1127
1128/**
1129 * skb_pad - zero pad the tail of an skb
1130 * @skb: buffer to pad
1131 * @pad: space to pad
1132 *
1133 * Ensure that a buffer is followed by a padding area that is zero
1134 * filled. Used by network drivers which may DMA or transfer data
1135 * beyond the buffer end onto the wire.
1136 *
1137 * May return error in out of memory cases. The skb is freed on error.
1138 */
1139static inline int skb_pad(struct sk_buff *skb, int pad)
1140{
1141 return __skb_pad(skb, pad, true);
1142}
ead2ceb0 1143#define dev_kfree_skb(a) consume_skb(a)
1da177e4 1144
be12a1fe
HFS
1145int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1146 int offset, size_t size);
1147
d94d9fee 1148struct skb_seq_state {
677e90ed
TG
1149 __u32 lower_offset;
1150 __u32 upper_offset;
1151 __u32 frag_idx;
1152 __u32 stepped_offset;
1153 struct sk_buff *root_skb;
1154 struct sk_buff *cur_skb;
1155 __u8 *frag_data;
1156};
1157
7965bd4d
JP
1158void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1159 unsigned int to, struct skb_seq_state *st);
1160unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1161 struct skb_seq_state *st);
1162void skb_abort_seq_read(struct skb_seq_state *st);
677e90ed 1163
7965bd4d 1164unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
059a2440 1165 unsigned int to, struct ts_config *config);
3fc7e8a6 1166
09323cc4
TH
1167/*
1168 * Packet hash types specify the type of hash in skb_set_hash.
1169 *
1170 * Hash types refer to the protocol layer addresses which are used to
1171 * construct a packet's hash. The hashes are used to differentiate or identify
1172 * flows of the protocol layer for the hash type. Hash types are either
1173 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1174 *
1175 * Properties of hashes:
1176 *
1177 * 1) Two packets in different flows have different hash values
1178 * 2) Two packets in the same flow should have the same hash value
1179 *
1180 * A hash at a higher layer is considered to be more specific. A driver should
1181 * set the most specific hash possible.
1182 *
1183 * A driver cannot indicate a more specific hash than the layer at which a hash
1184 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1185 *
1186 * A driver may indicate a hash level which is less specific than the
1187 * actual layer the hash was computed on. For instance, a hash computed
1188 * at L4 may be considered an L3 hash. This should only be done if the
1189 * driver can't unambiguously determine that the HW computed the hash at
1190 * the higher layer. Note that the "should" in the second property above
1191 * permits this.
1192 */
1193enum pkt_hash_types {
1194 PKT_HASH_TYPE_NONE, /* Undefined type */
1195 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1196 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1197 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1198};
1199
bcc83839 1200static inline void skb_clear_hash(struct sk_buff *skb)
09323cc4 1201{
bcc83839 1202 skb->hash = 0;
a3b18ddb 1203 skb->sw_hash = 0;
bcc83839
TH
1204 skb->l4_hash = 0;
1205}
1206
1207static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1208{
1209 if (!skb->l4_hash)
1210 skb_clear_hash(skb);
1211}
1212
1213static inline void
1214__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1215{
1216 skb->l4_hash = is_l4;
1217 skb->sw_hash = is_sw;
61b905da 1218 skb->hash = hash;
09323cc4
TH
1219}
1220
bcc83839
TH
1221static inline void
1222skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1223{
1224 /* Used by drivers to set hash from HW */
1225 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1226}
1227
1228static inline void
1229__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1230{
1231 __skb_set_hash(skb, hash, true, is_l4);
1232}
1233
e5276937 1234void __skb_get_hash(struct sk_buff *skb);
b917783c 1235u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
e5276937
TH
1236u32 skb_get_poff(const struct sk_buff *skb);
1237u32 __skb_get_poff(const struct sk_buff *skb, void *data,
72a338bc 1238 const struct flow_keys_basic *keys, int hlen);
e5276937
TH
1239__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1240 void *data, int hlen_proto);
1241
1242static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1243 int thoff, u8 ip_proto)
1244{
1245 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1246}
1247
1248void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1249 const struct flow_dissector_key *key,
1250 unsigned int key_count);
1251
2dfd184a 1252#ifdef CONFIG_NET
118c8e9a
SF
1253int skb_flow_dissector_prog_query(const union bpf_attr *attr,
1254 union bpf_attr __user *uattr);
d58e468b
PP
1255int skb_flow_dissector_bpf_prog_attach(const union bpf_attr *attr,
1256 struct bpf_prog *prog);
1257
1258int skb_flow_dissector_bpf_prog_detach(const union bpf_attr *attr);
2dfd184a 1259#else
118c8e9a
SF
1260static inline int skb_flow_dissector_prog_query(const union bpf_attr *attr,
1261 union bpf_attr __user *uattr)
1262{
1263 return -EOPNOTSUPP;
1264}
1265
2dfd184a
WB
1266static inline int skb_flow_dissector_bpf_prog_attach(const union bpf_attr *attr,
1267 struct bpf_prog *prog)
1268{
1269 return -EOPNOTSUPP;
1270}
1271
1272static inline int skb_flow_dissector_bpf_prog_detach(const union bpf_attr *attr)
1273{
1274 return -EOPNOTSUPP;
1275}
1276#endif
d58e468b 1277
089b19a9
SF
1278struct bpf_flow_dissector;
1279bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
086f9568 1280 __be16 proto, int nhoff, int hlen, unsigned int flags);
089b19a9 1281
3cbf4ffb
SF
1282bool __skb_flow_dissect(const struct net *net,
1283 const struct sk_buff *skb,
e5276937
TH
1284 struct flow_dissector *flow_dissector,
1285 void *target_container,
cd79a238
TH
1286 void *data, __be16 proto, int nhoff, int hlen,
1287 unsigned int flags);
e5276937
TH
1288
1289static inline bool skb_flow_dissect(const struct sk_buff *skb,
1290 struct flow_dissector *flow_dissector,
cd79a238 1291 void *target_container, unsigned int flags)
e5276937 1292{
3cbf4ffb
SF
1293 return __skb_flow_dissect(NULL, skb, flow_dissector,
1294 target_container, NULL, 0, 0, 0, flags);
e5276937
TH
1295}
1296
1297static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
cd79a238
TH
1298 struct flow_keys *flow,
1299 unsigned int flags)
e5276937
TH
1300{
1301 memset(flow, 0, sizeof(*flow));
3cbf4ffb
SF
1302 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1303 flow, NULL, 0, 0, 0, flags);
e5276937
TH
1304}
1305
72a338bc 1306static inline bool
3cbf4ffb
SF
1307skb_flow_dissect_flow_keys_basic(const struct net *net,
1308 const struct sk_buff *skb,
72a338bc
PA
1309 struct flow_keys_basic *flow, void *data,
1310 __be16 proto, int nhoff, int hlen,
1311 unsigned int flags)
e5276937
TH
1312{
1313 memset(flow, 0, sizeof(*flow));
3cbf4ffb 1314 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
cd79a238 1315 data, proto, nhoff, hlen, flags);
e5276937
TH
1316}
1317
82828b88
JP
1318void skb_flow_dissect_meta(const struct sk_buff *skb,
1319 struct flow_dissector *flow_dissector,
1320 void *target_container);
1321
75a56758
PB
1322/* Gets a skb connection tracking info, ctinfo map should be a
1323 * a map of mapsize to translate enum ip_conntrack_info states
1324 * to user states.
1325 */
1326void
1327skb_flow_dissect_ct(const struct sk_buff *skb,
1328 struct flow_dissector *flow_dissector,
1329 void *target_container,
1330 u16 *ctinfo_map,
1331 size_t mapsize);
62b32379
SH
1332void
1333skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1334 struct flow_dissector *flow_dissector,
1335 void *target_container);
1336
3958afa1 1337static inline __u32 skb_get_hash(struct sk_buff *skb)
bfb564e7 1338{
a3b18ddb 1339 if (!skb->l4_hash && !skb->sw_hash)
3958afa1 1340 __skb_get_hash(skb);
bfb564e7 1341
61b905da 1342 return skb->hash;
bfb564e7
KK
1343}
1344
20a17bf6 1345static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
f70ea018 1346{
c6cc1ca7
TH
1347 if (!skb->l4_hash && !skb->sw_hash) {
1348 struct flow_keys keys;
de4c1f8b 1349 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
c6cc1ca7 1350
de4c1f8b 1351 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
c6cc1ca7 1352 }
f70ea018
TH
1353
1354 return skb->hash;
1355}
1356
55667441
ED
1357__u32 skb_get_hash_perturb(const struct sk_buff *skb,
1358 const siphash_key_t *perturb);
50fb7992 1359
57bdf7f4
TH
1360static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1361{
61b905da 1362 return skb->hash;
57bdf7f4
TH
1363}
1364
3df7a74e
TH
1365static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1366{
61b905da 1367 to->hash = from->hash;
a3b18ddb 1368 to->sw_hash = from->sw_hash;
61b905da 1369 to->l4_hash = from->l4_hash;
3df7a74e
TH
1370};
1371
41477662
JK
1372static inline void skb_copy_decrypted(struct sk_buff *to,
1373 const struct sk_buff *from)
1374{
1375#ifdef CONFIG_TLS_DEVICE
1376 to->decrypted = from->decrypted;
1377#endif
1378}
1379
4305b541
ACM
1380#ifdef NET_SKBUFF_DATA_USES_OFFSET
1381static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1382{
1383 return skb->head + skb->end;
1384}
ec47ea82
AD
1385
1386static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1387{
1388 return skb->end;
1389}
4305b541
ACM
1390#else
1391static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1392{
1393 return skb->end;
1394}
ec47ea82
AD
1395
1396static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1397{
1398 return skb->end - skb->head;
1399}
4305b541
ACM
1400#endif
1401
1da177e4 1402/* Internal */
4305b541 1403#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
1da177e4 1404
ac45f602
PO
1405static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1406{
1407 return &skb_shinfo(skb)->hwtstamps;
1408}
1409
52267790
WB
1410static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1411{
1412 bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY;
1413
1414 return is_zcopy ? skb_uarg(skb) : NULL;
1415}
1416
52900d22
WB
1417static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1418 bool *have_ref)
52267790
WB
1419{
1420 if (skb && uarg && !skb_zcopy(skb)) {
52900d22
WB
1421 if (unlikely(have_ref && *have_ref))
1422 *have_ref = false;
1423 else
1424 sock_zerocopy_get(uarg);
52267790
WB
1425 skb_shinfo(skb)->destructor_arg = uarg;
1426 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1427 }
1428}
1429
5cd8d46e
WB
1430static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1431{
1432 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1433 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1434}
1435
1436static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1437{
1438 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1439}
1440
1441static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1442{
1443 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1444}
1445
52267790
WB
1446/* Release a reference on a zerocopy structure */
1447static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy)
1448{
1449 struct ubuf_info *uarg = skb_zcopy(skb);
1450
1451 if (uarg) {
185ce5c3
WB
1452 if (skb_zcopy_is_nouarg(skb)) {
1453 /* no notification callback */
1454 } else if (uarg->callback == sock_zerocopy_callback) {
0a4a060b
WB
1455 uarg->zerocopy = uarg->zerocopy && zerocopy;
1456 sock_zerocopy_put(uarg);
185ce5c3 1457 } else {
0a4a060b
WB
1458 uarg->callback(uarg, zerocopy);
1459 }
1460
52267790
WB
1461 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1462 }
1463}
1464
1465/* Abort a zerocopy operation and revert zckey on error in send syscall */
1466static inline void skb_zcopy_abort(struct sk_buff *skb)
1467{
1468 struct ubuf_info *uarg = skb_zcopy(skb);
1469
1470 if (uarg) {
52900d22 1471 sock_zerocopy_put_abort(uarg, false);
52267790
WB
1472 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1473 }
1474}
1475
a8305bff
DM
1476static inline void skb_mark_not_on_list(struct sk_buff *skb)
1477{
1478 skb->next = NULL;
1479}
1480
992cba7e
DM
1481static inline void skb_list_del_init(struct sk_buff *skb)
1482{
1483 __list_del_entry(&skb->list);
1484 skb_mark_not_on_list(skb);
1485}
1486
1da177e4
LT
1487/**
1488 * skb_queue_empty - check if a queue is empty
1489 * @list: queue head
1490 *
1491 * Returns true if the queue is empty, false otherwise.
1492 */
1493static inline int skb_queue_empty(const struct sk_buff_head *list)
1494{
fd44b93c 1495 return list->next == (const struct sk_buff *) list;
1da177e4
LT
1496}
1497
d7d16a89
ED
1498/**
1499 * skb_queue_empty_lockless - check if a queue is empty
1500 * @list: queue head
1501 *
1502 * Returns true if the queue is empty, false otherwise.
1503 * This variant can be used in lockless contexts.
1504 */
1505static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1506{
1507 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1508}
1509
1510
fc7ebb21
DM
1511/**
1512 * skb_queue_is_last - check if skb is the last entry in the queue
1513 * @list: queue head
1514 * @skb: buffer
1515 *
1516 * Returns true if @skb is the last buffer on the list.
1517 */
1518static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1519 const struct sk_buff *skb)
1520{
fd44b93c 1521 return skb->next == (const struct sk_buff *) list;
fc7ebb21
DM
1522}
1523
832d11c5
IJ
1524/**
1525 * skb_queue_is_first - check if skb is the first entry in the queue
1526 * @list: queue head
1527 * @skb: buffer
1528 *
1529 * Returns true if @skb is the first buffer on the list.
1530 */
1531static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1532 const struct sk_buff *skb)
1533{
fd44b93c 1534 return skb->prev == (const struct sk_buff *) list;
832d11c5
IJ
1535}
1536
249c8b42
DM
1537/**
1538 * skb_queue_next - return the next packet in the queue
1539 * @list: queue head
1540 * @skb: current buffer
1541 *
1542 * Return the next packet in @list after @skb. It is only valid to
1543 * call this if skb_queue_is_last() evaluates to false.
1544 */
1545static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1546 const struct sk_buff *skb)
1547{
1548 /* This BUG_ON may seem severe, but if we just return then we
1549 * are going to dereference garbage.
1550 */
1551 BUG_ON(skb_queue_is_last(list, skb));
1552 return skb->next;
1553}
1554
832d11c5
IJ
1555/**
1556 * skb_queue_prev - return the prev packet in the queue
1557 * @list: queue head
1558 * @skb: current buffer
1559 *
1560 * Return the prev packet in @list before @skb. It is only valid to
1561 * call this if skb_queue_is_first() evaluates to false.
1562 */
1563static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1564 const struct sk_buff *skb)
1565{
1566 /* This BUG_ON may seem severe, but if we just return then we
1567 * are going to dereference garbage.
1568 */
1569 BUG_ON(skb_queue_is_first(list, skb));
1570 return skb->prev;
1571}
1572
1da177e4
LT
1573/**
1574 * skb_get - reference buffer
1575 * @skb: buffer to reference
1576 *
1577 * Makes another reference to a socket buffer and returns a pointer
1578 * to the buffer.
1579 */
1580static inline struct sk_buff *skb_get(struct sk_buff *skb)
1581{
63354797 1582 refcount_inc(&skb->users);
1da177e4
LT
1583 return skb;
1584}
1585
1586/*
f8821f96 1587 * If users == 1, we are the only owner and can avoid redundant atomic changes.
1da177e4
LT
1588 */
1589
1da177e4
LT
1590/**
1591 * skb_cloned - is the buffer a clone
1592 * @skb: buffer to check
1593 *
1594 * Returns true if the buffer was generated with skb_clone() and is
1595 * one of multiple shared copies of the buffer. Cloned buffers are
1596 * shared data so must not be written to under normal circumstances.
1597 */
1598static inline int skb_cloned(const struct sk_buff *skb)
1599{
1600 return skb->cloned &&
1601 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1602}
1603
14bbd6a5
PS
1604static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1605{
d0164adc 1606 might_sleep_if(gfpflags_allow_blocking(pri));
14bbd6a5
PS
1607
1608 if (skb_cloned(skb))
1609 return pskb_expand_head(skb, 0, 0, pri);
1610
1611 return 0;
1612}
1613
1da177e4
LT
1614/**
1615 * skb_header_cloned - is the header a clone
1616 * @skb: buffer to check
1617 *
1618 * Returns true if modifying the header part of the buffer requires
1619 * the data to be copied.
1620 */
1621static inline int skb_header_cloned(const struct sk_buff *skb)
1622{
1623 int dataref;
1624
1625 if (!skb->cloned)
1626 return 0;
1627
1628 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1629 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1630 return dataref != 1;
1631}
1632
9580bf2e
ED
1633static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1634{
1635 might_sleep_if(gfpflags_allow_blocking(pri));
1636
1637 if (skb_header_cloned(skb))
1638 return pskb_expand_head(skb, 0, 0, pri);
1639
1640 return 0;
1641}
1642
f4a775d1
ED
1643/**
1644 * __skb_header_release - release reference to header
1645 * @skb: buffer to operate on
f4a775d1
ED
1646 */
1647static inline void __skb_header_release(struct sk_buff *skb)
1648{
1649 skb->nohdr = 1;
1650 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1651}
1652
1653
1da177e4
LT
1654/**
1655 * skb_shared - is the buffer shared
1656 * @skb: buffer to check
1657 *
1658 * Returns true if more than one person has a reference to this
1659 * buffer.
1660 */
1661static inline int skb_shared(const struct sk_buff *skb)
1662{
63354797 1663 return refcount_read(&skb->users) != 1;
1da177e4
LT
1664}
1665
1666/**
1667 * skb_share_check - check if buffer is shared and if so clone it
1668 * @skb: buffer to check
1669 * @pri: priority for memory allocation
1670 *
1671 * If the buffer is shared the buffer is cloned and the old copy
1672 * drops a reference. A new clone with a single reference is returned.
1673 * If the buffer is not shared the original buffer is returned. When
1674 * being called from interrupt status or with spinlocks held pri must
1675 * be GFP_ATOMIC.
1676 *
1677 * NULL is returned on a memory allocation failure.
1678 */
47061bc4 1679static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
1da177e4 1680{
d0164adc 1681 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1682 if (skb_shared(skb)) {
1683 struct sk_buff *nskb = skb_clone(skb, pri);
47061bc4
ED
1684
1685 if (likely(nskb))
1686 consume_skb(skb);
1687 else
1688 kfree_skb(skb);
1da177e4
LT
1689 skb = nskb;
1690 }
1691 return skb;
1692}
1693
1694/*
1695 * Copy shared buffers into a new sk_buff. We effectively do COW on
1696 * packets to handle cases where we have a local reader and forward
1697 * and a couple of other messy ones. The normal one is tcpdumping
1698 * a packet thats being forwarded.
1699 */
1700
1701/**
1702 * skb_unshare - make a copy of a shared buffer
1703 * @skb: buffer to check
1704 * @pri: priority for memory allocation
1705 *
1706 * If the socket buffer is a clone then this function creates a new
1707 * copy of the data, drops a reference count on the old copy and returns
1708 * the new copy with the reference count at 1. If the buffer is not a clone
1709 * the original buffer is returned. When called with a spinlock held or
1710 * from interrupt state @pri must be %GFP_ATOMIC
1711 *
1712 * %NULL is returned on a memory allocation failure.
1713 */
e2bf521d 1714static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
dd0fc66f 1715 gfp_t pri)
1da177e4 1716{
d0164adc 1717 might_sleep_if(gfpflags_allow_blocking(pri));
1da177e4
LT
1718 if (skb_cloned(skb)) {
1719 struct sk_buff *nskb = skb_copy(skb, pri);
31eff81e
AA
1720
1721 /* Free our shared copy */
1722 if (likely(nskb))
1723 consume_skb(skb);
1724 else
1725 kfree_skb(skb);
1da177e4
LT
1726 skb = nskb;
1727 }
1728 return skb;
1729}
1730
1731/**
1a5778aa 1732 * skb_peek - peek at the head of an &sk_buff_head
1da177e4
LT
1733 * @list_: list to peek at
1734 *
1735 * Peek an &sk_buff. Unlike most other operations you _MUST_
1736 * be careful with this one. A peek leaves the buffer on the
1737 * list and someone else may run off with it. You must hold
1738 * the appropriate locks or have a private queue to do this.
1739 *
1740 * Returns %NULL for an empty list or a pointer to the head element.
1741 * The reference count is not incremented and the reference is therefore
1742 * volatile. Use with caution.
1743 */
05bdd2f1 1744static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
1da177e4 1745{
18d07000
ED
1746 struct sk_buff *skb = list_->next;
1747
1748 if (skb == (struct sk_buff *)list_)
1749 skb = NULL;
1750 return skb;
1da177e4
LT
1751}
1752
8b69bd7d
DM
1753/**
1754 * __skb_peek - peek at the head of a non-empty &sk_buff_head
1755 * @list_: list to peek at
1756 *
1757 * Like skb_peek(), but the caller knows that the list is not empty.
1758 */
1759static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1760{
1761 return list_->next;
1762}
1763
da5ef6e5
PE
1764/**
1765 * skb_peek_next - peek skb following the given one from a queue
1766 * @skb: skb to start from
1767 * @list_: list to peek at
1768 *
1769 * Returns %NULL when the end of the list is met or a pointer to the
1770 * next element. The reference count is not incremented and the
1771 * reference is therefore volatile. Use with caution.
1772 */
1773static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1774 const struct sk_buff_head *list_)
1775{
1776 struct sk_buff *next = skb->next;
18d07000 1777
da5ef6e5
PE
1778 if (next == (struct sk_buff *)list_)
1779 next = NULL;
1780 return next;
1781}
1782
1da177e4 1783/**
1a5778aa 1784 * skb_peek_tail - peek at the tail of an &sk_buff_head
1da177e4
LT
1785 * @list_: list to peek at
1786 *
1787 * Peek an &sk_buff. Unlike most other operations you _MUST_
1788 * be careful with this one. A peek leaves the buffer on the
1789 * list and someone else may run off with it. You must hold
1790 * the appropriate locks or have a private queue to do this.
1791 *
1792 * Returns %NULL for an empty list or a pointer to the tail element.
1793 * The reference count is not incremented and the reference is therefore
1794 * volatile. Use with caution.
1795 */
05bdd2f1 1796static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
1da177e4 1797{
f8cc62ca 1798 struct sk_buff *skb = READ_ONCE(list_->prev);
18d07000
ED
1799
1800 if (skb == (struct sk_buff *)list_)
1801 skb = NULL;
1802 return skb;
1803
1da177e4
LT
1804}
1805
1806/**
1807 * skb_queue_len - get queue length
1808 * @list_: list to measure
1809 *
1810 * Return the length of an &sk_buff queue.
1811 */
1812static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1813{
1814 return list_->qlen;
1815}
1816
67fed459
DM
1817/**
1818 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1819 * @list: queue to initialize
1820 *
1821 * This initializes only the list and queue length aspects of
1822 * an sk_buff_head object. This allows to initialize the list
1823 * aspects of an sk_buff_head without reinitializing things like
1824 * the spinlock. It can also be used for on-stack sk_buff_head
1825 * objects where the spinlock is known to not be used.
1826 */
1827static inline void __skb_queue_head_init(struct sk_buff_head *list)
1828{
1829 list->prev = list->next = (struct sk_buff *)list;
1830 list->qlen = 0;
1831}
1832
76f10ad0
AV
1833/*
1834 * This function creates a split out lock class for each invocation;
1835 * this is needed for now since a whole lot of users of the skb-queue
1836 * infrastructure in drivers have different locking usage (in hardirq)
1837 * than the networking core (in softirq only). In the long run either the
1838 * network layer or drivers should need annotation to consolidate the
1839 * main types of usage into 3 classes.
1840 */
1da177e4
LT
1841static inline void skb_queue_head_init(struct sk_buff_head *list)
1842{
1843 spin_lock_init(&list->lock);
67fed459 1844 __skb_queue_head_init(list);
1da177e4
LT
1845}
1846
c2ecba71
PE
1847static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1848 struct lock_class_key *class)
1849{
1850 skb_queue_head_init(list);
1851 lockdep_set_class(&list->lock, class);
1852}
1853
1da177e4 1854/*
bf299275 1855 * Insert an sk_buff on a list.
1da177e4
LT
1856 *
1857 * The "__skb_xxxx()" functions are the non-atomic ones that
1858 * can only be called with interrupts disabled.
1859 */
bf299275
GR
1860static inline void __skb_insert(struct sk_buff *newsk,
1861 struct sk_buff *prev, struct sk_buff *next,
1862 struct sk_buff_head *list)
1863{
f8cc62ca
ED
1864 /* See skb_queue_empty_lockless() and skb_peek_tail()
1865 * for the opposite READ_ONCE()
1866 */
d7d16a89
ED
1867 WRITE_ONCE(newsk->next, next);
1868 WRITE_ONCE(newsk->prev, prev);
1869 WRITE_ONCE(next->prev, newsk);
1870 WRITE_ONCE(prev->next, newsk);
bf299275
GR
1871 list->qlen++;
1872}
1da177e4 1873
67fed459
DM
1874static inline void __skb_queue_splice(const struct sk_buff_head *list,
1875 struct sk_buff *prev,
1876 struct sk_buff *next)
1877{
1878 struct sk_buff *first = list->next;
1879 struct sk_buff *last = list->prev;
1880
d7d16a89
ED
1881 WRITE_ONCE(first->prev, prev);
1882 WRITE_ONCE(prev->next, first);
67fed459 1883
d7d16a89
ED
1884 WRITE_ONCE(last->next, next);
1885 WRITE_ONCE(next->prev, last);
67fed459
DM
1886}
1887
1888/**
1889 * skb_queue_splice - join two skb lists, this is designed for stacks
1890 * @list: the new list to add
1891 * @head: the place to add it in the first list
1892 */
1893static inline void skb_queue_splice(const struct sk_buff_head *list,
1894 struct sk_buff_head *head)
1895{
1896 if (!skb_queue_empty(list)) {
1897 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1898 head->qlen += list->qlen;
67fed459
DM
1899 }
1900}
1901
1902/**
d9619496 1903 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1904 * @list: the new list to add
1905 * @head: the place to add it in the first list
1906 *
1907 * The list at @list is reinitialised
1908 */
1909static inline void skb_queue_splice_init(struct sk_buff_head *list,
1910 struct sk_buff_head *head)
1911{
1912 if (!skb_queue_empty(list)) {
1913 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
1d4a31dd 1914 head->qlen += list->qlen;
67fed459
DM
1915 __skb_queue_head_init(list);
1916 }
1917}
1918
1919/**
1920 * skb_queue_splice_tail - join two skb lists, each list being a queue
1921 * @list: the new list to add
1922 * @head: the place to add it in the first list
1923 */
1924static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1925 struct sk_buff_head *head)
1926{
1927 if (!skb_queue_empty(list)) {
1928 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1929 head->qlen += list->qlen;
67fed459
DM
1930 }
1931}
1932
1933/**
d9619496 1934 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
67fed459
DM
1935 * @list: the new list to add
1936 * @head: the place to add it in the first list
1937 *
1938 * Each of the lists is a queue.
1939 * The list at @list is reinitialised
1940 */
1941static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1942 struct sk_buff_head *head)
1943{
1944 if (!skb_queue_empty(list)) {
1945 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
1d4a31dd 1946 head->qlen += list->qlen;
67fed459
DM
1947 __skb_queue_head_init(list);
1948 }
1949}
1950
1da177e4 1951/**
300ce174 1952 * __skb_queue_after - queue a buffer at the list head
1da177e4 1953 * @list: list to use
300ce174 1954 * @prev: place after this buffer
1da177e4
LT
1955 * @newsk: buffer to queue
1956 *
300ce174 1957 * Queue a buffer int the middle of a list. This function takes no locks
1da177e4
LT
1958 * and you must therefore hold required locks before calling it.
1959 *
1960 * A buffer cannot be placed on two lists at the same time.
1961 */
300ce174
SH
1962static inline void __skb_queue_after(struct sk_buff_head *list,
1963 struct sk_buff *prev,
1964 struct sk_buff *newsk)
1da177e4 1965{
bf299275 1966 __skb_insert(newsk, prev, prev->next, list);
1da177e4
LT
1967}
1968
7965bd4d
JP
1969void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1970 struct sk_buff_head *list);
7de6c033 1971
f5572855
GR
1972static inline void __skb_queue_before(struct sk_buff_head *list,
1973 struct sk_buff *next,
1974 struct sk_buff *newsk)
1975{
1976 __skb_insert(newsk, next->prev, next, list);
1977}
1978
300ce174
SH
1979/**
1980 * __skb_queue_head - queue a buffer at the list head
1981 * @list: list to use
1982 * @newsk: buffer to queue
1983 *
1984 * Queue a buffer at the start of a list. This function takes no locks
1985 * and you must therefore hold required locks before calling it.
1986 *
1987 * A buffer cannot be placed on two lists at the same time.
1988 */
300ce174
SH
1989static inline void __skb_queue_head(struct sk_buff_head *list,
1990 struct sk_buff *newsk)
1991{
1992 __skb_queue_after(list, (struct sk_buff *)list, newsk);
1993}
4ea7b0cf 1994void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
300ce174 1995
1da177e4
LT
1996/**
1997 * __skb_queue_tail - queue a buffer at the list tail
1998 * @list: list to use
1999 * @newsk: buffer to queue
2000 *
2001 * Queue a buffer at the end of a list. This function takes no locks
2002 * and you must therefore hold required locks before calling it.
2003 *
2004 * A buffer cannot be placed on two lists at the same time.
2005 */
1da177e4
LT
2006static inline void __skb_queue_tail(struct sk_buff_head *list,
2007 struct sk_buff *newsk)
2008{
f5572855 2009 __skb_queue_before(list, (struct sk_buff *)list, newsk);
1da177e4 2010}
4ea7b0cf 2011void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1da177e4 2012
1da177e4
LT
2013/*
2014 * remove sk_buff from list. _Must_ be called atomically, and with
2015 * the list known..
2016 */
7965bd4d 2017void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
1da177e4
LT
2018static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2019{
2020 struct sk_buff *next, *prev;
2021
2022 list->qlen--;
2023 next = skb->next;
2024 prev = skb->prev;
2025 skb->next = skb->prev = NULL;
d7d16a89
ED
2026 WRITE_ONCE(next->prev, prev);
2027 WRITE_ONCE(prev->next, next);
1da177e4
LT
2028}
2029
f525c06d
GR
2030/**
2031 * __skb_dequeue - remove from the head of the queue
2032 * @list: list to dequeue from
2033 *
2034 * Remove the head of the list. This function does not take any locks
2035 * so must be used with appropriate locks held only. The head item is
2036 * returned or %NULL if the list is empty.
2037 */
f525c06d
GR
2038static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2039{
2040 struct sk_buff *skb = skb_peek(list);
2041 if (skb)
2042 __skb_unlink(skb, list);
2043 return skb;
2044}
4ea7b0cf 2045struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1da177e4
LT
2046
2047/**
2048 * __skb_dequeue_tail - remove from the tail of the queue
2049 * @list: list to dequeue from
2050 *
2051 * Remove the tail of the list. This function does not take any locks
2052 * so must be used with appropriate locks held only. The tail item is
2053 * returned or %NULL if the list is empty.
2054 */
1da177e4
LT
2055static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2056{
2057 struct sk_buff *skb = skb_peek_tail(list);
2058 if (skb)
2059 __skb_unlink(skb, list);
2060 return skb;
2061}
4ea7b0cf 2062struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1da177e4
LT
2063
2064
bdcc0924 2065static inline bool skb_is_nonlinear(const struct sk_buff *skb)
1da177e4
LT
2066{
2067 return skb->data_len;
2068}
2069
2070static inline unsigned int skb_headlen(const struct sk_buff *skb)
2071{
2072 return skb->len - skb->data_len;
2073}
2074
3ece7826 2075static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
1da177e4 2076{
c72d8cda 2077 unsigned int i, len = 0;
1da177e4 2078
c72d8cda 2079 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
9e903e08 2080 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
3ece7826
WB
2081 return len;
2082}
2083
2084static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2085{
2086 return skb_headlen(skb) + __skb_pagelen(skb);
1da177e4
LT
2087}
2088
131ea667
IC
2089/**
2090 * __skb_fill_page_desc - initialise a paged fragment in an skb
2091 * @skb: buffer containing fragment to be initialised
2092 * @i: paged fragment index to initialise
2093 * @page: the page to use for this fragment
2094 * @off: the offset to the data with @page
2095 * @size: the length of the data
2096 *
2097 * Initialises the @i'th fragment of @skb to point to &size bytes at
2098 * offset @off within @page.
2099 *
2100 * Does not take any additional reference on the fragment.
2101 */
2102static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2103 struct page *page, int off, int size)
1da177e4
LT
2104{
2105 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2106
c48a11c7 2107 /*
2f064f34
MH
2108 * Propagate page pfmemalloc to the skb if we can. The problem is
2109 * that not all callers have unique ownership of the page but rely
2110 * on page_is_pfmemalloc doing the right thing(tm).
c48a11c7 2111 */
1dfa5bd3 2112 frag->bv_page = page;
65c84f14 2113 frag->bv_offset = off;
9e903e08 2114 skb_frag_size_set(frag, size);
cca7af38
PE
2115
2116 page = compound_head(page);
2f064f34 2117 if (page_is_pfmemalloc(page))
cca7af38 2118 skb->pfmemalloc = true;
131ea667
IC
2119}
2120
2121/**
2122 * skb_fill_page_desc - initialise a paged fragment in an skb
2123 * @skb: buffer containing fragment to be initialised
2124 * @i: paged fragment index to initialise
2125 * @page: the page to use for this fragment
2126 * @off: the offset to the data with @page
2127 * @size: the length of the data
2128 *
2129 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
bc32383c 2130 * @skb to point to @size bytes at offset @off within @page. In
131ea667
IC
2131 * addition updates @skb such that @i is the last fragment.
2132 *
2133 * Does not take any additional reference on the fragment.
2134 */
2135static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2136 struct page *page, int off, int size)
2137{
2138 __skb_fill_page_desc(skb, i, page, off, size);
1da177e4
LT
2139 skb_shinfo(skb)->nr_frags = i + 1;
2140}
2141
7965bd4d
JP
2142void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2143 int size, unsigned int truesize);
654bed16 2144
f8e617e1
JW
2145void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2146 unsigned int truesize);
2147
1da177e4
LT
2148#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2149
27a884dc
ACM
2150#ifdef NET_SKBUFF_DATA_USES_OFFSET
2151static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2152{
2153 return skb->head + skb->tail;
2154}
2155
2156static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2157{
2158 skb->tail = skb->data - skb->head;
2159}
2160
2161static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2162{
2163 skb_reset_tail_pointer(skb);
2164 skb->tail += offset;
2165}
7cc46190 2166
27a884dc
ACM
2167#else /* NET_SKBUFF_DATA_USES_OFFSET */
2168static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2169{
2170 return skb->tail;
2171}
2172
2173static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2174{
2175 skb->tail = skb->data;
2176}
2177
2178static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2179{
2180 skb->tail = skb->data + offset;
2181}
4305b541 2182
27a884dc
ACM
2183#endif /* NET_SKBUFF_DATA_USES_OFFSET */
2184
1da177e4
LT
2185/*
2186 * Add data to an sk_buff
2187 */
4df864c1
JB
2188void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2189void *skb_put(struct sk_buff *skb, unsigned int len);
2190static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
1da177e4 2191{
4df864c1 2192 void *tmp = skb_tail_pointer(skb);
1da177e4
LT
2193 SKB_LINEAR_ASSERT(skb);
2194 skb->tail += len;
2195 skb->len += len;
2196 return tmp;
2197}
2198
de77b966 2199static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2200{
2201 void *tmp = __skb_put(skb, len);
2202
2203 memset(tmp, 0, len);
2204 return tmp;
2205}
2206
2207static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2208 unsigned int len)
2209{
2210 void *tmp = __skb_put(skb, len);
2211
2212 memcpy(tmp, data, len);
2213 return tmp;
2214}
2215
2216static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2217{
2218 *(u8 *)__skb_put(skb, 1) = val;
2219}
2220
83ad357d 2221static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
e45a79da 2222{
83ad357d 2223 void *tmp = skb_put(skb, len);
e45a79da
JB
2224
2225 memset(tmp, 0, len);
2226
2227 return tmp;
2228}
2229
59ae1d12
JB
2230static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2231 unsigned int len)
2232{
2233 void *tmp = skb_put(skb, len);
2234
2235 memcpy(tmp, data, len);
2236
2237 return tmp;
2238}
2239
634fef61
JB
2240static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2241{
2242 *(u8 *)skb_put(skb, 1) = val;
2243}
2244
d58ff351
JB
2245void *skb_push(struct sk_buff *skb, unsigned int len);
2246static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2247{
2248 skb->data -= len;
2249 skb->len += len;
2250 return skb->data;
2251}
2252
af72868b
JB
2253void *skb_pull(struct sk_buff *skb, unsigned int len);
2254static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2255{
2256 skb->len -= len;
2257 BUG_ON(skb->len < skb->data_len);
2258 return skb->data += len;
2259}
2260
af72868b 2261static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
47d29646
DM
2262{
2263 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2264}
2265
af72868b 2266void *__pskb_pull_tail(struct sk_buff *skb, int delta);
1da177e4 2267
af72868b 2268static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2269{
2270 if (len > skb_headlen(skb) &&
987c402a 2271 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
1da177e4
LT
2272 return NULL;
2273 skb->len -= len;
2274 return skb->data += len;
2275}
2276
af72868b 2277static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2278{
2279 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2280}
2281
b9df4fd7 2282static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
1da177e4
LT
2283{
2284 if (likely(len <= skb_headlen(skb)))
b9df4fd7 2285 return true;
1da177e4 2286 if (unlikely(len > skb->len))
b9df4fd7 2287 return false;
987c402a 2288 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
1da177e4
LT
2289}
2290
c8c8b127
ED
2291void skb_condense(struct sk_buff *skb);
2292
1da177e4
LT
2293/**
2294 * skb_headroom - bytes at buffer head
2295 * @skb: buffer to check
2296 *
2297 * Return the number of bytes of free space at the head of an &sk_buff.
2298 */
c2636b4d 2299static inline unsigned int skb_headroom(const struct sk_buff *skb)
1da177e4
LT
2300{
2301 return skb->data - skb->head;
2302}
2303
2304/**
2305 * skb_tailroom - bytes at buffer end
2306 * @skb: buffer to check
2307 *
2308 * Return the number of bytes of free space at the tail of an sk_buff
2309 */
2310static inline int skb_tailroom(const struct sk_buff *skb)
2311{
4305b541 2312 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
1da177e4
LT
2313}
2314
a21d4572
ED
2315/**
2316 * skb_availroom - bytes at buffer end
2317 * @skb: buffer to check
2318 *
2319 * Return the number of bytes of free space at the tail of an sk_buff
2320 * allocated by sk_stream_alloc()
2321 */
2322static inline int skb_availroom(const struct sk_buff *skb)
2323{
16fad69c
ED
2324 if (skb_is_nonlinear(skb))
2325 return 0;
2326
2327 return skb->end - skb->tail - skb->reserved_tailroom;
a21d4572
ED
2328}
2329
1da177e4
LT
2330/**
2331 * skb_reserve - adjust headroom
2332 * @skb: buffer to alter
2333 * @len: bytes to move
2334 *
2335 * Increase the headroom of an empty &sk_buff by reducing the tail
2336 * room. This is only allowed for an empty buffer.
2337 */
8243126c 2338static inline void skb_reserve(struct sk_buff *skb, int len)
1da177e4
LT
2339{
2340 skb->data += len;
2341 skb->tail += len;
2342}
2343
1837b2e2
BP
2344/**
2345 * skb_tailroom_reserve - adjust reserved_tailroom
2346 * @skb: buffer to alter
2347 * @mtu: maximum amount of headlen permitted
2348 * @needed_tailroom: minimum amount of reserved_tailroom
2349 *
2350 * Set reserved_tailroom so that headlen can be as large as possible but
2351 * not larger than mtu and tailroom cannot be smaller than
2352 * needed_tailroom.
2353 * The required headroom should already have been reserved before using
2354 * this function.
2355 */
2356static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2357 unsigned int needed_tailroom)
2358{
2359 SKB_LINEAR_ASSERT(skb);
2360 if (mtu < skb_tailroom(skb) - needed_tailroom)
2361 /* use at most mtu */
2362 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2363 else
2364 /* use up to all available space */
2365 skb->reserved_tailroom = needed_tailroom;
2366}
2367
8bce6d7d
TH
2368#define ENCAP_TYPE_ETHER 0
2369#define ENCAP_TYPE_IPPROTO 1
2370
2371static inline void skb_set_inner_protocol(struct sk_buff *skb,
2372 __be16 protocol)
2373{
2374 skb->inner_protocol = protocol;
2375 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2376}
2377
2378static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2379 __u8 ipproto)
2380{
2381 skb->inner_ipproto = ipproto;
2382 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2383}
2384
6a674e9c
JG
2385static inline void skb_reset_inner_headers(struct sk_buff *skb)
2386{
aefbd2b3 2387 skb->inner_mac_header = skb->mac_header;
6a674e9c
JG
2388 skb->inner_network_header = skb->network_header;
2389 skb->inner_transport_header = skb->transport_header;
2390}
2391
0b5c9db1
JP
2392static inline void skb_reset_mac_len(struct sk_buff *skb)
2393{
2394 skb->mac_len = skb->network_header - skb->mac_header;
2395}
2396
6a674e9c
JG
2397static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2398 *skb)
2399{
2400 return skb->head + skb->inner_transport_header;
2401}
2402
55dc5a9f
TH
2403static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2404{
2405 return skb_inner_transport_header(skb) - skb->data;
2406}
2407
6a674e9c
JG
2408static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2409{
2410 skb->inner_transport_header = skb->data - skb->head;
2411}
2412
2413static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2414 const int offset)
2415{
2416 skb_reset_inner_transport_header(skb);
2417 skb->inner_transport_header += offset;
2418}
2419
2420static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2421{
2422 return skb->head + skb->inner_network_header;
2423}
2424
2425static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2426{
2427 skb->inner_network_header = skb->data - skb->head;
2428}
2429
2430static inline void skb_set_inner_network_header(struct sk_buff *skb,
2431 const int offset)
2432{
2433 skb_reset_inner_network_header(skb);
2434 skb->inner_network_header += offset;
2435}
2436
aefbd2b3
PS
2437static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2438{
2439 return skb->head + skb->inner_mac_header;
2440}
2441
2442static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2443{
2444 skb->inner_mac_header = skb->data - skb->head;
2445}
2446
2447static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2448 const int offset)
2449{
2450 skb_reset_inner_mac_header(skb);
2451 skb->inner_mac_header += offset;
2452}
fda55eca
ED
2453static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2454{
35d04610 2455 return skb->transport_header != (typeof(skb->transport_header))~0U;
fda55eca
ED
2456}
2457
9c70220b
ACM
2458static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2459{
2e07fa9c 2460 return skb->head + skb->transport_header;
9c70220b
ACM
2461}
2462
badff6d0
ACM
2463static inline void skb_reset_transport_header(struct sk_buff *skb)
2464{
2e07fa9c 2465 skb->transport_header = skb->data - skb->head;
badff6d0
ACM
2466}
2467
967b05f6
ACM
2468static inline void skb_set_transport_header(struct sk_buff *skb,
2469 const int offset)
2470{
2e07fa9c
ACM
2471 skb_reset_transport_header(skb);
2472 skb->transport_header += offset;
ea2ae17d
ACM
2473}
2474
d56f90a7
ACM
2475static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2476{
2e07fa9c 2477 return skb->head + skb->network_header;
d56f90a7
ACM
2478}
2479
c1d2bbe1
ACM
2480static inline void skb_reset_network_header(struct sk_buff *skb)
2481{
2e07fa9c 2482 skb->network_header = skb->data - skb->head;
c1d2bbe1
ACM
2483}
2484
c14d2450
ACM
2485static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2486{
2e07fa9c
ACM
2487 skb_reset_network_header(skb);
2488 skb->network_header += offset;
c14d2450
ACM
2489}
2490
2e07fa9c 2491static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
bbe735e4 2492{
2e07fa9c 2493 return skb->head + skb->mac_header;
bbe735e4
ACM
2494}
2495
ea6da4fd
AV
2496static inline int skb_mac_offset(const struct sk_buff *skb)
2497{
2498 return skb_mac_header(skb) - skb->data;
2499}
2500
0daf4349
DB
2501static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2502{
2503 return skb->network_header - skb->mac_header;
2504}
2505
2e07fa9c 2506static inline int skb_mac_header_was_set(const struct sk_buff *skb)
cfe1fc77 2507{
35d04610 2508 return skb->mac_header != (typeof(skb->mac_header))~0U;
2e07fa9c
ACM
2509}
2510
2511static inline void skb_reset_mac_header(struct sk_buff *skb)
2512{
2513 skb->mac_header = skb->data - skb->head;
2514}
2515
2516static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2517{
2518 skb_reset_mac_header(skb);
2519 skb->mac_header += offset;
2520}
2521
0e3da5bb
TT
2522static inline void skb_pop_mac_header(struct sk_buff *skb)
2523{
2524 skb->mac_header = skb->network_header;
2525}
2526
d2aa125d 2527static inline void skb_probe_transport_header(struct sk_buff *skb)
fbbdb8f0 2528{
72a338bc 2529 struct flow_keys_basic keys;
fbbdb8f0
YX
2530
2531 if (skb_transport_header_was_set(skb))
2532 return;
72a338bc 2533
3cbf4ffb
SF
2534 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2535 NULL, 0, 0, 0, 0))
42aecaa9 2536 skb_set_transport_header(skb, keys.control.thoff);
fbbdb8f0
YX
2537}
2538
03606895
ED
2539static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2540{
2541 if (skb_mac_header_was_set(skb)) {
2542 const unsigned char *old_mac = skb_mac_header(skb);
2543
2544 skb_set_mac_header(skb, -skb->mac_len);
2545 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2546 }
2547}
2548
04fb451e
MM
2549static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2550{
2551 return skb->csum_start - skb_headroom(skb);
2552}
2553
08b64fcc
AD
2554static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2555{
2556 return skb->head + skb->csum_start;
2557}
2558
2e07fa9c
ACM
2559static inline int skb_transport_offset(const struct sk_buff *skb)
2560{
2561 return skb_transport_header(skb) - skb->data;
2562}
2563
2564static inline u32 skb_network_header_len(const struct sk_buff *skb)
2565{
2566 return skb->transport_header - skb->network_header;
2567}
2568
6a674e9c
JG
2569static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2570{
2571 return skb->inner_transport_header - skb->inner_network_header;
2572}
2573
2e07fa9c
ACM
2574static inline int skb_network_offset(const struct sk_buff *skb)
2575{
2576 return skb_network_header(skb) - skb->data;
2577}
48d49d0c 2578
6a674e9c
JG
2579static inline int skb_inner_network_offset(const struct sk_buff *skb)
2580{
2581 return skb_inner_network_header(skb) - skb->data;
2582}
2583
f9599ce1
CG
2584static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2585{
2586 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2587}
2588
1da177e4
LT
2589/*
2590 * CPUs often take a performance hit when accessing unaligned memory
2591 * locations. The actual performance hit varies, it can be small if the
2592 * hardware handles it or large if we have to take an exception and fix it
2593 * in software.
2594 *
2595 * Since an ethernet header is 14 bytes network drivers often end up with
2596 * the IP header at an unaligned offset. The IP header can be aligned by
2597 * shifting the start of the packet by 2 bytes. Drivers should do this
2598 * with:
2599 *
8660c124 2600 * skb_reserve(skb, NET_IP_ALIGN);
1da177e4
LT
2601 *
2602 * The downside to this alignment of the IP header is that the DMA is now
2603 * unaligned. On some architectures the cost of an unaligned DMA is high
2604 * and this cost outweighs the gains made by aligning the IP header.
8660c124 2605 *
1da177e4
LT
2606 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2607 * to be overridden.
2608 */
2609#ifndef NET_IP_ALIGN
2610#define NET_IP_ALIGN 2
2611#endif
2612
025be81e
AB
2613/*
2614 * The networking layer reserves some headroom in skb data (via
2615 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2616 * the header has to grow. In the default case, if the header has to grow
d6301d3d 2617 * 32 bytes or less we avoid the reallocation.
025be81e
AB
2618 *
2619 * Unfortunately this headroom changes the DMA alignment of the resulting
2620 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2621 * on some architectures. An architecture can override this value,
2622 * perhaps setting it to a cacheline in size (since that will maintain
2623 * cacheline alignment of the DMA). It must be a power of 2.
2624 *
d6301d3d 2625 * Various parts of the networking layer expect at least 32 bytes of
025be81e 2626 * headroom, you should not reduce this.
5933dd2f
ED
2627 *
2628 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2629 * to reduce average number of cache lines per packet.
2630 * get_rps_cpus() for example only access one 64 bytes aligned block :
18e8c134 2631 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
025be81e
AB
2632 */
2633#ifndef NET_SKB_PAD
5933dd2f 2634#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
025be81e
AB
2635#endif
2636
7965bd4d 2637int ___pskb_trim(struct sk_buff *skb, unsigned int len);
1da177e4 2638
5293efe6 2639static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
1da177e4 2640{
5e1abdc3 2641 if (WARN_ON(skb_is_nonlinear(skb)))
3cc0e873 2642 return;
27a884dc
ACM
2643 skb->len = len;
2644 skb_set_tail_pointer(skb, len);
1da177e4
LT
2645}
2646
5293efe6
DB
2647static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2648{
2649 __skb_set_length(skb, len);
2650}
2651
7965bd4d 2652void skb_trim(struct sk_buff *skb, unsigned int len);
1da177e4
LT
2653
2654static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2655{
3cc0e873
HX
2656 if (skb->data_len)
2657 return ___pskb_trim(skb, len);
2658 __skb_trim(skb, len);
2659 return 0;
1da177e4
LT
2660}
2661
2662static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2663{
2664 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2665}
2666
e9fa4f7b
HX
2667/**
2668 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2669 * @skb: buffer to alter
2670 * @len: new length
2671 *
2672 * This is identical to pskb_trim except that the caller knows that
2673 * the skb is not cloned so we should never get an error due to out-
2674 * of-memory.
2675 */
2676static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2677{
2678 int err = pskb_trim(skb, len);
2679 BUG_ON(err);
2680}
2681
5293efe6
DB
2682static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2683{
2684 unsigned int diff = len - skb->len;
2685
2686 if (skb_tailroom(skb) < diff) {
2687 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2688 GFP_ATOMIC);
2689 if (ret)
2690 return ret;
2691 }
2692 __skb_set_length(skb, len);
2693 return 0;
2694}
2695
1da177e4
LT
2696/**
2697 * skb_orphan - orphan a buffer
2698 * @skb: buffer to orphan
2699 *
2700 * If a buffer currently has an owner then we call the owner's
2701 * destructor function and make the @skb unowned. The buffer continues
2702 * to exist but is no longer charged to its former owner.
2703 */
2704static inline void skb_orphan(struct sk_buff *skb)
2705{
c34a7612 2706 if (skb->destructor) {
1da177e4 2707 skb->destructor(skb);
c34a7612
ED
2708 skb->destructor = NULL;
2709 skb->sk = NULL;
376c7311
ED
2710 } else {
2711 BUG_ON(skb->sk);
c34a7612 2712 }
1da177e4
LT
2713}
2714
a353e0ce
MT
2715/**
2716 * skb_orphan_frags - orphan the frags contained in a buffer
2717 * @skb: buffer to orphan frags from
2718 * @gfp_mask: allocation mask for replacement pages
2719 *
2720 * For each frag in the SKB which needs a destructor (i.e. has an
2721 * owner) create a copy of that frag and release the original
2722 * page by calling the destructor.
2723 */
2724static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2725{
1f8b977a
WB
2726 if (likely(!skb_zcopy(skb)))
2727 return 0;
185ce5c3
WB
2728 if (!skb_zcopy_is_nouarg(skb) &&
2729 skb_uarg(skb)->callback == sock_zerocopy_callback)
1f8b977a
WB
2730 return 0;
2731 return skb_copy_ubufs(skb, gfp_mask);
2732}
2733
2734/* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2735static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2736{
2737 if (likely(!skb_zcopy(skb)))
a353e0ce
MT
2738 return 0;
2739 return skb_copy_ubufs(skb, gfp_mask);
2740}
2741
1da177e4
LT
2742/**
2743 * __skb_queue_purge - empty a list
2744 * @list: list to empty
2745 *
2746 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2747 * the list and one reference dropped. This function does not take the
2748 * list lock and the caller must hold the relevant locks to use it.
2749 */
1da177e4
LT
2750static inline void __skb_queue_purge(struct sk_buff_head *list)
2751{
2752 struct sk_buff *skb;
2753 while ((skb = __skb_dequeue(list)) != NULL)
2754 kfree_skb(skb);
2755}
4ea7b0cf 2756void skb_queue_purge(struct sk_buff_head *list);
1da177e4 2757
385114de 2758unsigned int skb_rbtree_purge(struct rb_root *root);
9f5afeae 2759
7965bd4d 2760void *netdev_alloc_frag(unsigned int fragsz);
1da177e4 2761
7965bd4d
JP
2762struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2763 gfp_t gfp_mask);
8af27456
CH
2764
2765/**
2766 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2767 * @dev: network device to receive on
2768 * @length: length to allocate
2769 *
2770 * Allocate a new &sk_buff and assign it a usage count of one. The
2771 * buffer has unspecified headroom built in. Users should allocate
2772 * the headroom they think they need without accounting for the
2773 * built in space. The built in space is used for optimisations.
2774 *
2775 * %NULL is returned if there is no free memory. Although this function
2776 * allocates memory it can be called from an interrupt.
2777 */
2778static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
6f532612 2779 unsigned int length)
8af27456
CH
2780{
2781 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2782}
2783
6f532612
ED
2784/* legacy helper around __netdev_alloc_skb() */
2785static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2786 gfp_t gfp_mask)
2787{
2788 return __netdev_alloc_skb(NULL, length, gfp_mask);
2789}
2790
2791/* legacy helper around netdev_alloc_skb() */
2792static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2793{
2794 return netdev_alloc_skb(NULL, length);
2795}
2796
2797
4915a0de
ED
2798static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2799 unsigned int length, gfp_t gfp)
61321bbd 2800{
4915a0de 2801 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
61321bbd
ED
2802
2803 if (NET_IP_ALIGN && skb)
2804 skb_reserve(skb, NET_IP_ALIGN);
2805 return skb;
2806}
2807
4915a0de
ED
2808static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2809 unsigned int length)
2810{
2811 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2812}
2813
181edb2b
AD
2814static inline void skb_free_frag(void *addr)
2815{
8c2dd3e4 2816 page_frag_free(addr);
181edb2b
AD
2817}
2818
ffde7328 2819void *napi_alloc_frag(unsigned int fragsz);
fd11a83d
AD
2820struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2821 unsigned int length, gfp_t gfp_mask);
2822static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2823 unsigned int length)
2824{
2825 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2826}
795bb1c0
JDB
2827void napi_consume_skb(struct sk_buff *skb, int budget);
2828
2829void __kfree_skb_flush(void);
15fad714 2830void __kfree_skb_defer(struct sk_buff *skb);
ffde7328 2831
71dfda58
AD
2832/**
2833 * __dev_alloc_pages - allocate page for network Rx
2834 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2835 * @order: size of the allocation
2836 *
2837 * Allocate a new page.
2838 *
2839 * %NULL is returned if there is no free memory.
2840*/
2841static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2842 unsigned int order)
2843{
2844 /* This piece of code contains several assumptions.
2845 * 1. This is for device Rx, therefor a cold page is preferred.
2846 * 2. The expectation is the user wants a compound page.
2847 * 3. If requesting a order 0 page it will not be compound
2848 * due to the check to see if order has a value in prep_new_page
2849 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2850 * code in gfp_to_alloc_flags that should be enforcing this.
2851 */
453f85d4 2852 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
71dfda58
AD
2853
2854 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2855}
2856
2857static inline struct page *dev_alloc_pages(unsigned int order)
2858{
95829b3a 2859 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
71dfda58
AD
2860}
2861
2862/**
2863 * __dev_alloc_page - allocate a page for network Rx
2864 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2865 *
2866 * Allocate a new page.
2867 *
2868 * %NULL is returned if there is no free memory.
2869 */
2870static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2871{
2872 return __dev_alloc_pages(gfp_mask, 0);
2873}
2874
2875static inline struct page *dev_alloc_page(void)
2876{
95829b3a 2877 return dev_alloc_pages(0);
71dfda58
AD
2878}
2879
0614002b
MG
2880/**
2881 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2882 * @page: The page that was allocated from skb_alloc_page
2883 * @skb: The skb that may need pfmemalloc set
2884 */
2885static inline void skb_propagate_pfmemalloc(struct page *page,
2886 struct sk_buff *skb)
2887{
2f064f34 2888 if (page_is_pfmemalloc(page))
0614002b
MG
2889 skb->pfmemalloc = true;
2890}
2891
7240b60c
JL
2892/**
2893 * skb_frag_off() - Returns the offset of a skb fragment
2894 * @frag: the paged fragment
2895 */
2896static inline unsigned int skb_frag_off(const skb_frag_t *frag)
2897{
65c84f14 2898 return frag->bv_offset;
7240b60c
JL
2899}
2900
2901/**
2902 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
2903 * @frag: skb fragment
2904 * @delta: value to add
2905 */
2906static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
2907{
65c84f14 2908 frag->bv_offset += delta;
7240b60c
JL
2909}
2910
2911/**
2912 * skb_frag_off_set() - Sets the offset of a skb fragment
2913 * @frag: skb fragment
2914 * @offset: offset of fragment
2915 */
2916static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
2917{
65c84f14 2918 frag->bv_offset = offset;
7240b60c
JL
2919}
2920
2921/**
2922 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
2923 * @fragto: skb fragment where offset is set
2924 * @fragfrom: skb fragment offset is copied from
2925 */
2926static inline void skb_frag_off_copy(skb_frag_t *fragto,
2927 const skb_frag_t *fragfrom)
2928{
65c84f14 2929 fragto->bv_offset = fragfrom->bv_offset;
7240b60c
JL
2930}
2931
131ea667 2932/**
e227867f 2933 * skb_frag_page - retrieve the page referred to by a paged fragment
131ea667
IC
2934 * @frag: the paged fragment
2935 *
2936 * Returns the &struct page associated with @frag.
2937 */
2938static inline struct page *skb_frag_page(const skb_frag_t *frag)
2939{
1dfa5bd3 2940 return frag->bv_page;
131ea667
IC
2941}
2942
2943/**
2944 * __skb_frag_ref - take an addition reference on a paged fragment.
2945 * @frag: the paged fragment
2946 *
2947 * Takes an additional reference on the paged fragment @frag.
2948 */
2949static inline void __skb_frag_ref(skb_frag_t *frag)
2950{
2951 get_page(skb_frag_page(frag));
2952}
2953
2954/**
2955 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2956 * @skb: the buffer
2957 * @f: the fragment offset.
2958 *
2959 * Takes an additional reference on the @f'th paged fragment of @skb.
2960 */
2961static inline void skb_frag_ref(struct sk_buff *skb, int f)
2962{
2963 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2964}
2965
2966/**
2967 * __skb_frag_unref - release a reference on a paged fragment.
2968 * @frag: the paged fragment
2969 *
2970 * Releases a reference on the paged fragment @frag.
2971 */
2972static inline void __skb_frag_unref(skb_frag_t *frag)
2973{
2974 put_page(skb_frag_page(frag));
2975}
2976
2977/**
2978 * skb_frag_unref - release a reference on a paged fragment of an skb.
2979 * @skb: the buffer
2980 * @f: the fragment offset
2981 *
2982 * Releases a reference on the @f'th paged fragment of @skb.
2983 */
2984static inline void skb_frag_unref(struct sk_buff *skb, int f)
2985{
2986 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2987}
2988
2989/**
2990 * skb_frag_address - gets the address of the data contained in a paged fragment
2991 * @frag: the paged fragment buffer
2992 *
2993 * Returns the address of the data within @frag. The page must already
2994 * be mapped.
2995 */
2996static inline void *skb_frag_address(const skb_frag_t *frag)
2997{
7240b60c 2998 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
131ea667
IC
2999}
3000
3001/**
3002 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3003 * @frag: the paged fragment buffer
3004 *
3005 * Returns the address of the data within @frag. Checks that the page
3006 * is mapped and returns %NULL otherwise.
3007 */
3008static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3009{
3010 void *ptr = page_address(skb_frag_page(frag));
3011 if (unlikely(!ptr))
3012 return NULL;
3013
7240b60c
JL
3014 return ptr + skb_frag_off(frag);
3015}
3016
3017/**
3018 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3019 * @fragto: skb fragment where page is set
3020 * @fragfrom: skb fragment page is copied from
3021 */
3022static inline void skb_frag_page_copy(skb_frag_t *fragto,
3023 const skb_frag_t *fragfrom)
3024{
3025 fragto->bv_page = fragfrom->bv_page;
131ea667
IC
3026}
3027
3028/**
3029 * __skb_frag_set_page - sets the page contained in a paged fragment
3030 * @frag: the paged fragment
3031 * @page: the page to set
3032 *
3033 * Sets the fragment @frag to contain @page.
3034 */
3035static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3036{
1dfa5bd3 3037 frag->bv_page = page;
131ea667
IC
3038}
3039
3040/**
3041 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3042 * @skb: the buffer
3043 * @f: the fragment offset
3044 * @page: the page to set
3045 *
3046 * Sets the @f'th fragment of @skb to contain @page.
3047 */
3048static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3049 struct page *page)
3050{
3051 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3052}
3053
400dfd3a
ED
3054bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3055
131ea667
IC
3056/**
3057 * skb_frag_dma_map - maps a paged fragment via the DMA API
f83347df 3058 * @dev: the device to map the fragment to
131ea667
IC
3059 * @frag: the paged fragment to map
3060 * @offset: the offset within the fragment (starting at the
3061 * fragment's own offset)
3062 * @size: the number of bytes to map
771b00a8 3063 * @dir: the direction of the mapping (``PCI_DMA_*``)
131ea667
IC
3064 *
3065 * Maps the page associated with @frag to @device.
3066 */
3067static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3068 const skb_frag_t *frag,
3069 size_t offset, size_t size,
3070 enum dma_data_direction dir)
3071{
3072 return dma_map_page(dev, skb_frag_page(frag),
7240b60c 3073 skb_frag_off(frag) + offset, size, dir);
131ea667
IC
3074}
3075
117632e6
ED
3076static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3077 gfp_t gfp_mask)
3078{
3079 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3080}
3081
bad93e9d
OP
3082
3083static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3084 gfp_t gfp_mask)
3085{
3086 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3087}
3088
3089
334a8132
PM
3090/**
3091 * skb_clone_writable - is the header of a clone writable
3092 * @skb: buffer to check
3093 * @len: length up to which to write
3094 *
3095 * Returns true if modifying the header part of the cloned buffer
3096 * does not requires the data to be copied.
3097 */
05bdd2f1 3098static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
334a8132
PM
3099{
3100 return !skb_header_cloned(skb) &&
3101 skb_headroom(skb) + len <= skb->hdr_len;
3102}
3103
3697649f
DB
3104static inline int skb_try_make_writable(struct sk_buff *skb,
3105 unsigned int write_len)
3106{
3107 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3108 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3109}
3110
d9cc2048
HX
3111static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3112 int cloned)
3113{
3114 int delta = 0;
3115
d9cc2048
HX
3116 if (headroom > skb_headroom(skb))
3117 delta = headroom - skb_headroom(skb);
3118
3119 if (delta || cloned)
3120 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3121 GFP_ATOMIC);
3122 return 0;
3123}
3124
1da177e4
LT
3125/**
3126 * skb_cow - copy header of skb when it is required
3127 * @skb: buffer to cow
3128 * @headroom: needed headroom
3129 *
3130 * If the skb passed lacks sufficient headroom or its data part
3131 * is shared, data is reallocated. If reallocation fails, an error
3132 * is returned and original skb is not changed.
3133 *
3134 * The result is skb with writable area skb->head...skb->tail
3135 * and at least @headroom of space at head.
3136 */
3137static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3138{
d9cc2048
HX
3139 return __skb_cow(skb, headroom, skb_cloned(skb));
3140}
1da177e4 3141
d9cc2048
HX
3142/**
3143 * skb_cow_head - skb_cow but only making the head writable
3144 * @skb: buffer to cow
3145 * @headroom: needed headroom
3146 *
3147 * This function is identical to skb_cow except that we replace the
3148 * skb_cloned check by skb_header_cloned. It should be used when
3149 * you only need to push on some header and do not need to modify
3150 * the data.
3151 */
3152static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3153{
3154 return __skb_cow(skb, headroom, skb_header_cloned(skb));
1da177e4
LT
3155}
3156
3157/**
3158 * skb_padto - pad an skbuff up to a minimal size
3159 * @skb: buffer to pad
3160 * @len: minimal length
3161 *
3162 * Pads up a buffer to ensure the trailing bytes exist and are
3163 * blanked. If the buffer already contains sufficient data it
5b057c6b
HX
3164 * is untouched. Otherwise it is extended. Returns zero on
3165 * success. The skb is freed on error.
1da177e4 3166 */
5b057c6b 3167static inline int skb_padto(struct sk_buff *skb, unsigned int len)
1da177e4
LT
3168{
3169 unsigned int size = skb->len;
3170 if (likely(size >= len))
5b057c6b 3171 return 0;
987c402a 3172 return skb_pad(skb, len - size);
1da177e4
LT
3173}
3174
9c0c1124 3175/**
4ea7b0cf 3176 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
9c0c1124
AD
3177 * @skb: buffer to pad
3178 * @len: minimal length
cd0a137a 3179 * @free_on_error: free buffer on error
9c0c1124
AD
3180 *
3181 * Pads up a buffer to ensure the trailing bytes exist and are
3182 * blanked. If the buffer already contains sufficient data it
3183 * is untouched. Otherwise it is extended. Returns zero on
cd0a137a 3184 * success. The skb is freed on error if @free_on_error is true.
9c0c1124 3185 */
cd0a137a
FF
3186static inline int __skb_put_padto(struct sk_buff *skb, unsigned int len,
3187 bool free_on_error)
9c0c1124
AD
3188{
3189 unsigned int size = skb->len;
3190
3191 if (unlikely(size < len)) {
3192 len -= size;
cd0a137a 3193 if (__skb_pad(skb, len, free_on_error))
9c0c1124
AD
3194 return -ENOMEM;
3195 __skb_put(skb, len);
3196 }
3197 return 0;
3198}
3199
cd0a137a
FF
3200/**
3201 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3202 * @skb: buffer to pad
3203 * @len: minimal length
3204 *
3205 * Pads up a buffer to ensure the trailing bytes exist and are
3206 * blanked. If the buffer already contains sufficient data it
3207 * is untouched. Otherwise it is extended. Returns zero on
3208 * success. The skb is freed on error.
3209 */
3210static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
3211{
3212 return __skb_put_padto(skb, len, true);
3213}
3214
1da177e4 3215static inline int skb_add_data(struct sk_buff *skb,
af2b040e 3216 struct iov_iter *from, int copy)
1da177e4
LT
3217{
3218 const int off = skb->len;
3219
3220 if (skb->ip_summed == CHECKSUM_NONE) {
af2b040e 3221 __wsum csum = 0;
15e6cb46
AV
3222 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3223 &csum, from)) {
1da177e4
LT
3224 skb->csum = csum_block_add(skb->csum, csum, off);
3225 return 0;
3226 }
15e6cb46 3227 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
1da177e4
LT
3228 return 0;
3229
3230 __skb_trim(skb, off);
3231 return -EFAULT;
3232}
3233
38ba0a65
ED
3234static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3235 const struct page *page, int off)
1da177e4 3236{
1f8b977a
WB
3237 if (skb_zcopy(skb))
3238 return false;
1da177e4 3239 if (i) {
d8e18a51 3240 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
1da177e4 3241
ea2ab693 3242 return page == skb_frag_page(frag) &&
7240b60c 3243 off == skb_frag_off(frag) + skb_frag_size(frag);
1da177e4 3244 }
38ba0a65 3245 return false;
1da177e4
LT
3246}
3247
364c6bad
HX
3248static inline int __skb_linearize(struct sk_buff *skb)
3249{
3250 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3251}
3252
1da177e4
LT
3253/**
3254 * skb_linearize - convert paged skb to linear one
3255 * @skb: buffer to linarize
1da177e4
LT
3256 *
3257 * If there is no free memory -ENOMEM is returned, otherwise zero
3258 * is returned and the old skb data released.
3259 */
364c6bad
HX
3260static inline int skb_linearize(struct sk_buff *skb)
3261{
3262 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3263}
3264
cef401de
ED
3265/**
3266 * skb_has_shared_frag - can any frag be overwritten
3267 * @skb: buffer to test
3268 *
3269 * Return true if the skb has at least one frag that might be modified
3270 * by an external entity (as in vmsplice()/sendfile())
3271 */
3272static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3273{
c9af6db4
PS
3274 return skb_is_nonlinear(skb) &&
3275 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
cef401de
ED
3276}
3277
364c6bad
HX
3278/**
3279 * skb_linearize_cow - make sure skb is linear and writable
3280 * @skb: buffer to process
3281 *
3282 * If there is no free memory -ENOMEM is returned, otherwise zero
3283 * is returned and the old skb data released.
3284 */
3285static inline int skb_linearize_cow(struct sk_buff *skb)
1da177e4 3286{
364c6bad
HX
3287 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3288 __skb_linearize(skb) : 0;
1da177e4
LT
3289}
3290
479ffccc
DB
3291static __always_inline void
3292__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3293 unsigned int off)
3294{
3295 if (skb->ip_summed == CHECKSUM_COMPLETE)
3296 skb->csum = csum_block_sub(skb->csum,
3297 csum_partial(start, len, 0), off);
3298 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3299 skb_checksum_start_offset(skb) < 0)
3300 skb->ip_summed = CHECKSUM_NONE;
3301}
3302
1da177e4
LT
3303/**
3304 * skb_postpull_rcsum - update checksum for received skb after pull
3305 * @skb: buffer to update
3306 * @start: start of data before pull
3307 * @len: length of data pulled
3308 *
3309 * After doing a pull on a received packet, you need to call this to
84fa7933
PM
3310 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3311 * CHECKSUM_NONE so that it can be recomputed from scratch.
1da177e4 3312 */
1da177e4 3313static inline void skb_postpull_rcsum(struct sk_buff *skb,
cbb042f9 3314 const void *start, unsigned int len)
1da177e4 3315{
479ffccc 3316 __skb_postpull_rcsum(skb, start, len, 0);
1da177e4
LT
3317}
3318
479ffccc
DB
3319static __always_inline void
3320__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3321 unsigned int off)
3322{
3323 if (skb->ip_summed == CHECKSUM_COMPLETE)
3324 skb->csum = csum_block_add(skb->csum,
3325 csum_partial(start, len, 0), off);
3326}
cbb042f9 3327
479ffccc
DB
3328/**
3329 * skb_postpush_rcsum - update checksum for received skb after push
3330 * @skb: buffer to update
3331 * @start: start of data after push
3332 * @len: length of data pushed
3333 *
3334 * After doing a push on a received packet, you need to call this to
3335 * update the CHECKSUM_COMPLETE checksum.
3336 */
f8ffad69
DB
3337static inline void skb_postpush_rcsum(struct sk_buff *skb,
3338 const void *start, unsigned int len)
3339{
479ffccc 3340 __skb_postpush_rcsum(skb, start, len, 0);
f8ffad69
DB
3341}
3342
af72868b 3343void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
479ffccc 3344
82a31b92
WC
3345/**
3346 * skb_push_rcsum - push skb and update receive checksum
3347 * @skb: buffer to update
3348 * @len: length of data pulled
3349 *
3350 * This function performs an skb_push on the packet and updates
3351 * the CHECKSUM_COMPLETE checksum. It should be used on
3352 * receive path processing instead of skb_push unless you know
3353 * that the checksum difference is zero (e.g., a valid IP header)
3354 * or you are setting ip_summed to CHECKSUM_NONE.
3355 */
d58ff351 3356static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
82a31b92
WC
3357{
3358 skb_push(skb, len);
3359 skb_postpush_rcsum(skb, skb->data, len);
3360 return skb->data;
3361}
3362
88078d98 3363int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
7ce5a27f
DM
3364/**
3365 * pskb_trim_rcsum - trim received skb and update checksum
3366 * @skb: buffer to trim
3367 * @len: new length
3368 *
3369 * This is exactly the same as pskb_trim except that it ensures the
3370 * checksum of received packets are still valid after the operation.
6c57f045 3371 * It can change skb pointers.
7ce5a27f
DM
3372 */
3373
3374static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3375{
3376 if (likely(len >= skb->len))
3377 return 0;
88078d98 3378 return pskb_trim_rcsum_slow(skb, len);
7ce5a27f
DM
3379}
3380
5293efe6
DB
3381static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3382{
3383 if (skb->ip_summed == CHECKSUM_COMPLETE)
3384 skb->ip_summed = CHECKSUM_NONE;
3385 __skb_trim(skb, len);
3386 return 0;
3387}
3388
3389static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3390{
3391 if (skb->ip_summed == CHECKSUM_COMPLETE)
3392 skb->ip_summed = CHECKSUM_NONE;
3393 return __skb_grow(skb, len);
3394}
3395
18a4c0ea
ED
3396#define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3397#define skb_rb_first(root) rb_to_skb(rb_first(root))
3398#define skb_rb_last(root) rb_to_skb(rb_last(root))
3399#define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3400#define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3401
1da177e4
LT
3402#define skb_queue_walk(queue, skb) \
3403 for (skb = (queue)->next; \
a1e4891f 3404 skb != (struct sk_buff *)(queue); \
1da177e4
LT
3405 skb = skb->next)
3406
46f8914e
JC
3407#define skb_queue_walk_safe(queue, skb, tmp) \
3408 for (skb = (queue)->next, tmp = skb->next; \
3409 skb != (struct sk_buff *)(queue); \
3410 skb = tmp, tmp = skb->next)
3411
1164f52a 3412#define skb_queue_walk_from(queue, skb) \
a1e4891f 3413 for (; skb != (struct sk_buff *)(queue); \
1164f52a
DM
3414 skb = skb->next)
3415
18a4c0ea
ED
3416#define skb_rbtree_walk(skb, root) \
3417 for (skb = skb_rb_first(root); skb != NULL; \
3418 skb = skb_rb_next(skb))
3419
3420#define skb_rbtree_walk_from(skb) \
3421 for (; skb != NULL; \
3422 skb = skb_rb_next(skb))
3423
3424#define skb_rbtree_walk_from_safe(skb, tmp) \
3425 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3426 skb = tmp)
3427
1164f52a
DM
3428#define skb_queue_walk_from_safe(queue, skb, tmp) \
3429 for (tmp = skb->next; \
3430 skb != (struct sk_buff *)(queue); \
3431 skb = tmp, tmp = skb->next)
3432
300ce174
SH
3433#define skb_queue_reverse_walk(queue, skb) \
3434 for (skb = (queue)->prev; \
a1e4891f 3435 skb != (struct sk_buff *)(queue); \
300ce174
SH
3436 skb = skb->prev)
3437
686a2955
DM
3438#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3439 for (skb = (queue)->prev, tmp = skb->prev; \
3440 skb != (struct sk_buff *)(queue); \
3441 skb = tmp, tmp = skb->prev)
3442
3443#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3444 for (tmp = skb->prev; \
3445 skb != (struct sk_buff *)(queue); \
3446 skb = tmp, tmp = skb->prev)
1da177e4 3447
21dc3301 3448static inline bool skb_has_frag_list(const struct sk_buff *skb)
ee039871
DM
3449{
3450 return skb_shinfo(skb)->frag_list != NULL;
3451}
3452
3453static inline void skb_frag_list_init(struct sk_buff *skb)
3454{
3455 skb_shinfo(skb)->frag_list = NULL;
3456}
3457
ee039871
DM
3458#define skb_walk_frags(skb, iter) \
3459 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3460
ea3793ee
RW
3461
3462int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
3463 const struct sk_buff *skb);
65101aec
PA
3464struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3465 struct sk_buff_head *queue,
3466 unsigned int flags,
3467 void (*destructor)(struct sock *sk,
3468 struct sk_buff *skb),
fd69c399 3469 int *off, int *err,
65101aec 3470 struct sk_buff **last);
ea3793ee 3471struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
7c13f97f
PA
3472 void (*destructor)(struct sock *sk,
3473 struct sk_buff *skb),
fd69c399 3474 int *off, int *err,
ea3793ee 3475 struct sk_buff **last);
7965bd4d 3476struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
7c13f97f
PA
3477 void (*destructor)(struct sock *sk,
3478 struct sk_buff *skb),
fd69c399 3479 int *off, int *err);
7965bd4d
JP
3480struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3481 int *err);
a11e1d43
LT
3482__poll_t datagram_poll(struct file *file, struct socket *sock,
3483 struct poll_table_struct *wait);
c0371da6
AV
3484int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3485 struct iov_iter *to, int size);
51f3d02b
DM
3486static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3487 struct msghdr *msg, int size)
3488{
e5a4b0bb 3489 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
227158db 3490}
e5a4b0bb
AV
3491int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3492 struct msghdr *msg);
65d69e25
SG
3493int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3494 struct iov_iter *to, int len,
3495 struct ahash_request *hash);
3a654f97
AV
3496int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3497 struct iov_iter *from, int len);
3a654f97 3498int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
7965bd4d 3499void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
627d2d6b 3500void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3501static inline void skb_free_datagram_locked(struct sock *sk,
3502 struct sk_buff *skb)
3503{
3504 __skb_free_datagram_locked(sk, skb, 0);
3505}
7965bd4d 3506int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
7965bd4d
JP
3507int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3508int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3509__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3510 int len, __wsum csum);
a60e3cc7 3511int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
7965bd4d 3512 struct pipe_inode_info *pipe, unsigned int len,
25869262 3513 unsigned int flags);
20bf50de
TH
3514int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3515 int len);
7965bd4d 3516void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
af2806f8 3517unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
36d5fe6a
ZK
3518int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3519 int len, int hlen);
7965bd4d
JP
3520void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3521int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3522void skb_scrub_packet(struct sk_buff *skb, bool xnet);
779b7931 3523bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
2b16f048 3524bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
7965bd4d 3525struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
0d5501c1 3526struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
e2195121 3527int skb_ensure_writable(struct sk_buff *skb, int write_len);
bfca4c52 3528int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
93515d53
JP
3529int skb_vlan_pop(struct sk_buff *skb);
3530int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
fa4e0f88
DC
3531int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
3532 int mac_len);
3533int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len);
d27cf5c5 3534int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
2a2ea508 3535int skb_mpls_dec_ttl(struct sk_buff *skb);
6fa01ccd
SV
3536struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3537 gfp_t gfp);
20380731 3538
6ce8e9ce
AV
3539static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3540{
3073f070 3541 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
6ce8e9ce
AV
3542}
3543
7eab8d9e
AV
3544static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3545{
e5a4b0bb 3546 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
7eab8d9e
AV
3547}
3548
2817a336
DB
3549struct skb_checksum_ops {
3550 __wsum (*update)(const void *mem, int len, __wsum wsum);
3551 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3552};
3553
9617813d
DC
3554extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3555
2817a336
DB
3556__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3557 __wsum csum, const struct skb_checksum_ops *ops);
3558__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3559 __wsum csum);
3560
1e98a0f0
ED
3561static inline void * __must_check
3562__skb_header_pointer(const struct sk_buff *skb, int offset,
3563 int len, void *data, int hlen, void *buffer)
1da177e4 3564{
55820ee2 3565 if (hlen - offset >= len)
690e36e7 3566 return data + offset;
1da177e4 3567
690e36e7
DM
3568 if (!skb ||
3569 skb_copy_bits(skb, offset, buffer, len) < 0)
1da177e4
LT
3570 return NULL;
3571
3572 return buffer;
3573}
3574
1e98a0f0
ED
3575static inline void * __must_check
3576skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
690e36e7
DM
3577{
3578 return __skb_header_pointer(skb, offset, len, skb->data,
3579 skb_headlen(skb), buffer);
3580}
3581
4262e5cc
DB
3582/**
3583 * skb_needs_linearize - check if we need to linearize a given skb
3584 * depending on the given device features.
3585 * @skb: socket buffer to check
3586 * @features: net device features
3587 *
3588 * Returns true if either:
3589 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3590 * 2. skb is fragmented and the device does not support SG.
3591 */
3592static inline bool skb_needs_linearize(struct sk_buff *skb,
3593 netdev_features_t features)
3594{
3595 return skb_is_nonlinear(skb) &&
3596 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3597 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3598}
3599
d626f62b
ACM
3600static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3601 void *to,
3602 const unsigned int len)
3603{
3604 memcpy(to, skb->data, len);
3605}
3606
3607static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3608 const int offset, void *to,
3609 const unsigned int len)
3610{
3611 memcpy(to, skb->data + offset, len);
3612}
3613
27d7ff46
ACM
3614static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3615 const void *from,
3616 const unsigned int len)
3617{
3618 memcpy(skb->data, from, len);
3619}
3620
3621static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3622 const int offset,
3623 const void *from,
3624 const unsigned int len)
3625{
3626 memcpy(skb->data + offset, from, len);
3627}
3628
7965bd4d 3629void skb_init(void);
1da177e4 3630
ac45f602
PO
3631static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3632{
3633 return skb->tstamp;
3634}
3635
a61bbcf2
PM
3636/**
3637 * skb_get_timestamp - get timestamp from a skb
3638 * @skb: skb to get stamp from
13c6ee2a 3639 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
a61bbcf2
PM
3640 *
3641 * Timestamps are stored in the skb as offsets to a base timestamp.
3642 * This function converts the offset back to a struct timeval and stores
3643 * it in stamp.
3644 */
ac45f602 3645static inline void skb_get_timestamp(const struct sk_buff *skb,
13c6ee2a 3646 struct __kernel_old_timeval *stamp)
a61bbcf2 3647{
13c6ee2a 3648 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
a61bbcf2
PM
3649}
3650
887feae3
DD
3651static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3652 struct __kernel_sock_timeval *stamp)
3653{
3654 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3655
3656 stamp->tv_sec = ts.tv_sec;
3657 stamp->tv_usec = ts.tv_nsec / 1000;
3658}
3659
ac45f602
PO
3660static inline void skb_get_timestampns(const struct sk_buff *skb,
3661 struct timespec *stamp)
3662{
3663 *stamp = ktime_to_timespec(skb->tstamp);
3664}
3665
887feae3
DD
3666static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3667 struct __kernel_timespec *stamp)
3668{
3669 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3670
3671 stamp->tv_sec = ts.tv_sec;
3672 stamp->tv_nsec = ts.tv_nsec;
3673}
3674
b7aa0bf7 3675static inline void __net_timestamp(struct sk_buff *skb)
a61bbcf2 3676{
b7aa0bf7 3677 skb->tstamp = ktime_get_real();
a61bbcf2
PM
3678}
3679
164891aa
SH
3680static inline ktime_t net_timedelta(ktime_t t)
3681{
3682 return ktime_sub(ktime_get_real(), t);
3683}
3684
b9ce204f
IJ
3685static inline ktime_t net_invalid_timestamp(void)
3686{
8b0e1953 3687 return 0;
b9ce204f 3688}
a61bbcf2 3689
de8f3a83
DB
3690static inline u8 skb_metadata_len(const struct sk_buff *skb)
3691{
3692 return skb_shinfo(skb)->meta_len;
3693}
3694
3695static inline void *skb_metadata_end(const struct sk_buff *skb)
3696{
3697 return skb_mac_header(skb);
3698}
3699
3700static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3701 const struct sk_buff *skb_b,
3702 u8 meta_len)
3703{
3704 const void *a = skb_metadata_end(skb_a);
3705 const void *b = skb_metadata_end(skb_b);
3706 /* Using more efficient varaiant than plain call to memcmp(). */
3707#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3708 u64 diffs = 0;
3709
3710 switch (meta_len) {
3711#define __it(x, op) (x -= sizeof(u##op))
3712#define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3713 case 32: diffs |= __it_diff(a, b, 64);
82385b0d 3714 /* fall through */
de8f3a83 3715 case 24: diffs |= __it_diff(a, b, 64);
82385b0d 3716 /* fall through */
de8f3a83 3717 case 16: diffs |= __it_diff(a, b, 64);
82385b0d 3718 /* fall through */
de8f3a83
DB
3719 case 8: diffs |= __it_diff(a, b, 64);
3720 break;
3721 case 28: diffs |= __it_diff(a, b, 64);
82385b0d 3722 /* fall through */
de8f3a83 3723 case 20: diffs |= __it_diff(a, b, 64);
82385b0d 3724 /* fall through */
de8f3a83 3725 case 12: diffs |= __it_diff(a, b, 64);
82385b0d 3726 /* fall through */
de8f3a83
DB
3727 case 4: diffs |= __it_diff(a, b, 32);
3728 break;
3729 }
3730 return diffs;
3731#else
3732 return memcmp(a - meta_len, b - meta_len, meta_len);
3733#endif
3734}
3735
3736static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3737 const struct sk_buff *skb_b)
3738{
3739 u8 len_a = skb_metadata_len(skb_a);
3740 u8 len_b = skb_metadata_len(skb_b);
3741
3742 if (!(len_a | len_b))
3743 return false;
3744
3745 return len_a != len_b ?
3746 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3747}
3748
3749static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3750{
3751 skb_shinfo(skb)->meta_len = meta_len;
3752}
3753
3754static inline void skb_metadata_clear(struct sk_buff *skb)
3755{
3756 skb_metadata_set(skb, 0);
3757}
3758
62bccb8c
AD
3759struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3760
c1f19b51
RC
3761#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3762
7965bd4d
JP
3763void skb_clone_tx_timestamp(struct sk_buff *skb);
3764bool skb_defer_rx_timestamp(struct sk_buff *skb);
c1f19b51
RC
3765
3766#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3767
3768static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3769{
3770}
3771
3772static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3773{
3774 return false;
3775}
3776
3777#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3778
3779/**
3780 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3781 *
da92b194
RC
3782 * PHY drivers may accept clones of transmitted packets for
3783 * timestamping via their phy_driver.txtstamp method. These drivers
7a76a021
BP
3784 * must call this function to return the skb back to the stack with a
3785 * timestamp.
da92b194 3786 *
c1f19b51 3787 * @skb: clone of the the original outgoing packet
7a76a021 3788 * @hwtstamps: hardware time stamps
c1f19b51
RC
3789 *
3790 */
3791void skb_complete_tx_timestamp(struct sk_buff *skb,
3792 struct skb_shared_hwtstamps *hwtstamps);
3793
e7fd2885
WB
3794void __skb_tstamp_tx(struct sk_buff *orig_skb,
3795 struct skb_shared_hwtstamps *hwtstamps,
3796 struct sock *sk, int tstype);
3797
ac45f602
PO
3798/**
3799 * skb_tstamp_tx - queue clone of skb with send time stamps
3800 * @orig_skb: the original outgoing packet
3801 * @hwtstamps: hardware time stamps, may be NULL if not available
3802 *
3803 * If the skb has a socket associated, then this function clones the
3804 * skb (thus sharing the actual data and optional structures), stores
3805 * the optional hardware time stamping information (if non NULL) or
3806 * generates a software time stamp (otherwise), then queues the clone
3807 * to the error queue of the socket. Errors are silently ignored.
3808 */
7965bd4d
JP
3809void skb_tstamp_tx(struct sk_buff *orig_skb,
3810 struct skb_shared_hwtstamps *hwtstamps);
ac45f602 3811
4507a715
RC
3812/**
3813 * skb_tx_timestamp() - Driver hook for transmit timestamping
3814 *
3815 * Ethernet MAC Drivers should call this function in their hard_xmit()
4ff75b7c 3816 * function immediately before giving the sk_buff to the MAC hardware.
4507a715 3817 *
73409f3b
DM
3818 * Specifically, one should make absolutely sure that this function is
3819 * called before TX completion of this packet can trigger. Otherwise
3820 * the packet could potentially already be freed.
3821 *
4507a715
RC
3822 * @skb: A socket buffer.
3823 */
3824static inline void skb_tx_timestamp(struct sk_buff *skb)
3825{
c1f19b51 3826 skb_clone_tx_timestamp(skb);
b50a5c70
ML
3827 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3828 skb_tstamp_tx(skb, NULL);
4507a715
RC
3829}
3830
6e3e939f
JB
3831/**
3832 * skb_complete_wifi_ack - deliver skb with wifi status
3833 *
3834 * @skb: the original outgoing packet
3835 * @acked: ack status
3836 *
3837 */
3838void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3839
7965bd4d
JP
3840__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3841__sum16 __skb_checksum_complete(struct sk_buff *skb);
fb286bb2 3842
60476372
HX
3843static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3844{
6edec0e6
TH
3845 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3846 skb->csum_valid ||
3847 (skb->ip_summed == CHECKSUM_PARTIAL &&
3848 skb_checksum_start_offset(skb) >= 0));
60476372
HX
3849}
3850
fb286bb2
HX
3851/**
3852 * skb_checksum_complete - Calculate checksum of an entire packet
3853 * @skb: packet to process
3854 *
3855 * This function calculates the checksum over the entire packet plus
3856 * the value of skb->csum. The latter can be used to supply the
3857 * checksum of a pseudo header as used by TCP/UDP. It returns the
3858 * checksum.
3859 *
3860 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3861 * this function can be used to verify that checksum on received
3862 * packets. In that case the function should return zero if the
3863 * checksum is correct. In particular, this function will return zero
3864 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3865 * hardware has already verified the correctness of the checksum.
3866 */
4381ca3c 3867static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
fb286bb2 3868{
60476372
HX
3869 return skb_csum_unnecessary(skb) ?
3870 0 : __skb_checksum_complete(skb);
fb286bb2
HX
3871}
3872
77cffe23
TH
3873static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3874{
3875 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3876 if (skb->csum_level == 0)
3877 skb->ip_summed = CHECKSUM_NONE;
3878 else
3879 skb->csum_level--;
3880 }
3881}
3882
3883static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3884{
3885 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3886 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3887 skb->csum_level++;
3888 } else if (skb->ip_summed == CHECKSUM_NONE) {
3889 skb->ip_summed = CHECKSUM_UNNECESSARY;
3890 skb->csum_level = 0;
3891 }
3892}
3893
76ba0aae
TH
3894/* Check if we need to perform checksum complete validation.
3895 *
3896 * Returns true if checksum complete is needed, false otherwise
3897 * (either checksum is unnecessary or zero checksum is allowed).
3898 */
3899static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3900 bool zero_okay,
3901 __sum16 check)
3902{
5d0c2b95
TH
3903 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3904 skb->csum_valid = 1;
77cffe23 3905 __skb_decr_checksum_unnecessary(skb);
76ba0aae
TH
3906 return false;
3907 }
3908
3909 return true;
3910}
3911
da279887 3912/* For small packets <= CHECKSUM_BREAK perform checksum complete directly
76ba0aae
TH
3913 * in checksum_init.
3914 */
3915#define CHECKSUM_BREAK 76
3916
4e18b9ad
TH
3917/* Unset checksum-complete
3918 *
3919 * Unset checksum complete can be done when packet is being modified
3920 * (uncompressed for instance) and checksum-complete value is
3921 * invalidated.
3922 */
3923static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3924{
3925 if (skb->ip_summed == CHECKSUM_COMPLETE)
3926 skb->ip_summed = CHECKSUM_NONE;
3927}
3928
76ba0aae
TH
3929/* Validate (init) checksum based on checksum complete.
3930 *
3931 * Return values:
3932 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3933 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3934 * checksum is stored in skb->csum for use in __skb_checksum_complete
3935 * non-zero: value of invalid checksum
3936 *
3937 */
3938static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3939 bool complete,
3940 __wsum psum)
3941{
3942 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3943 if (!csum_fold(csum_add(psum, skb->csum))) {
5d0c2b95 3944 skb->csum_valid = 1;
76ba0aae
TH
3945 return 0;
3946 }
3947 }
3948
3949 skb->csum = psum;
3950
5d0c2b95
TH
3951 if (complete || skb->len <= CHECKSUM_BREAK) {
3952 __sum16 csum;
3953
3954 csum = __skb_checksum_complete(skb);
3955 skb->csum_valid = !csum;
3956 return csum;
3957 }
76ba0aae
TH
3958
3959 return 0;
3960}
3961
3962static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3963{
3964 return 0;
3965}
3966
3967/* Perform checksum validate (init). Note that this is a macro since we only
3968 * want to calculate the pseudo header which is an input function if necessary.
3969 * First we try to validate without any computation (checksum unnecessary) and
3970 * then calculate based on checksum complete calling the function to compute
3971 * pseudo header.
3972 *
3973 * Return values:
3974 * 0: checksum is validated or try to in skb_checksum_complete
3975 * non-zero: value of invalid checksum
3976 */
3977#define __skb_checksum_validate(skb, proto, complete, \
3978 zero_okay, check, compute_pseudo) \
3979({ \
3980 __sum16 __ret = 0; \
5d0c2b95 3981 skb->csum_valid = 0; \
76ba0aae
TH
3982 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3983 __ret = __skb_checksum_validate_complete(skb, \
3984 complete, compute_pseudo(skb, proto)); \
3985 __ret; \
3986})
3987
3988#define skb_checksum_init(skb, proto, compute_pseudo) \
3989 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3990
3991#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3992 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3993
3994#define skb_checksum_validate(skb, proto, compute_pseudo) \
3995 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3996
3997#define skb_checksum_validate_zero_check(skb, proto, check, \
3998 compute_pseudo) \
096a4cfa 3999 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
76ba0aae
TH
4000
4001#define skb_checksum_simple_validate(skb) \
4002 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4003
d96535a1
TH
4004static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4005{
219f1d79 4006 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
d96535a1
TH
4007}
4008
e4aa33ad 4009static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
d96535a1
TH
4010{
4011 skb->csum = ~pseudo;
4012 skb->ip_summed = CHECKSUM_COMPLETE;
4013}
4014
e4aa33ad 4015#define skb_checksum_try_convert(skb, proto, compute_pseudo) \
d96535a1
TH
4016do { \
4017 if (__skb_checksum_convert_check(skb)) \
e4aa33ad 4018 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
d96535a1
TH
4019} while (0)
4020
15e2396d
TH
4021static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4022 u16 start, u16 offset)
4023{
4024 skb->ip_summed = CHECKSUM_PARTIAL;
4025 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4026 skb->csum_offset = offset - start;
4027}
4028
dcdc8994
TH
4029/* Update skbuf and packet to reflect the remote checksum offload operation.
4030 * When called, ptr indicates the starting point for skb->csum when
4031 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4032 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4033 */
4034static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
15e2396d 4035 int start, int offset, bool nopartial)
dcdc8994
TH
4036{
4037 __wsum delta;
4038
15e2396d
TH
4039 if (!nopartial) {
4040 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4041 return;
4042 }
4043
dcdc8994
TH
4044 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4045 __skb_checksum_complete(skb);
4046 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4047 }
4048
4049 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4050
4051 /* Adjust skb->csum since we changed the packet */
4052 skb->csum = csum_add(skb->csum, delta);
4053}
4054
cb9c6836
FW
4055static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4056{
4057#if IS_ENABLED(CONFIG_NF_CONNTRACK)
261db6c2 4058 return (void *)(skb->_nfct & NFCT_PTRMASK);
cb9c6836
FW
4059#else
4060 return NULL;
4061#endif
4062}
4063
261db6c2 4064static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
1da177e4 4065{
261db6c2
JS
4066#if IS_ENABLED(CONFIG_NF_CONNTRACK)
4067 return skb->_nfct;
4068#else
4069 return 0UL;
4070#endif
1da177e4 4071}
261db6c2
JS
4072
4073static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
1da177e4 4074{
261db6c2
JS
4075#if IS_ENABLED(CONFIG_NF_CONNTRACK)
4076 skb->_nfct = nfct;
2fc72c7b 4077#endif
261db6c2 4078}
df5042f4
FW
4079
4080#ifdef CONFIG_SKB_EXTENSIONS
4081enum skb_ext_id {
4082#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4083 SKB_EXT_BRIDGE_NF,
4165079b
FW
4084#endif
4085#ifdef CONFIG_XFRM
4086 SKB_EXT_SEC_PATH,
95a7233c
PB
4087#endif
4088#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4089 TC_SKB_EXT,
df5042f4
FW
4090#endif
4091 SKB_EXT_NUM, /* must be last */
4092};
4093
4094/**
4095 * struct skb_ext - sk_buff extensions
4096 * @refcnt: 1 on allocation, deallocated on 0
4097 * @offset: offset to add to @data to obtain extension address
4098 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4099 * @data: start of extension data, variable sized
4100 *
4101 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4102 * to use 'u8' types while allowing up to 2kb worth of extension data.
4103 */
4104struct skb_ext {
4105 refcount_t refcnt;
4106 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4107 u8 chunks; /* same */
4108 char data[0] __aligned(8);
4109};
4110
4111void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4112void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4113void __skb_ext_put(struct skb_ext *ext);
4114
4115static inline void skb_ext_put(struct sk_buff *skb)
4116{
4117 if (skb->active_extensions)
4118 __skb_ext_put(skb->extensions);
4119}
4120
df5042f4
FW
4121static inline void __skb_ext_copy(struct sk_buff *dst,
4122 const struct sk_buff *src)
4123{
4124 dst->active_extensions = src->active_extensions;
4125
4126 if (src->active_extensions) {
4127 struct skb_ext *ext = src->extensions;
4128
4129 refcount_inc(&ext->refcnt);
4130 dst->extensions = ext;
4131 }
4132}
4133
4134static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4135{
4136 skb_ext_put(dst);
4137 __skb_ext_copy(dst, src);
4138}
4139
4140static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4141{
4142 return !!ext->offset[i];
4143}
4144
4145static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4146{
4147 return skb->active_extensions & (1 << id);
4148}
4149
4150static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4151{
4152 if (skb_ext_exist(skb, id))
4153 __skb_ext_del(skb, id);
4154}
4155
4156static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4157{
4158 if (skb_ext_exist(skb, id)) {
4159 struct skb_ext *ext = skb->extensions;
4160
4161 return (void *)ext + (ext->offset[id] << 3);
4162 }
4163
4164 return NULL;
4165}
174e2381
FW
4166
4167static inline void skb_ext_reset(struct sk_buff *skb)
4168{
4169 if (unlikely(skb->active_extensions)) {
4170 __skb_ext_put(skb->extensions);
4171 skb->active_extensions = 0;
4172 }
4173}
677bf08c
FW
4174
4175static inline bool skb_has_extensions(struct sk_buff *skb)
4176{
4177 return unlikely(skb->active_extensions);
4178}
df5042f4
FW
4179#else
4180static inline void skb_ext_put(struct sk_buff *skb) {}
174e2381 4181static inline void skb_ext_reset(struct sk_buff *skb) {}
df5042f4
FW
4182static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
4183static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
4184static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
677bf08c 4185static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
df5042f4
FW
4186#endif /* CONFIG_SKB_EXTENSIONS */
4187
895b5c9f 4188static inline void nf_reset_ct(struct sk_buff *skb)
a193a4ab 4189{
5f79e0f9 4190#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
4191 nf_conntrack_put(skb_nfct(skb));
4192 skb->_nfct = 0;
2fc72c7b 4193#endif
a193a4ab
PM
4194}
4195
124dff01
PM
4196static inline void nf_reset_trace(struct sk_buff *skb)
4197{
478b360a 4198#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
130549fe
G
4199 skb->nf_trace = 0;
4200#endif
a193a4ab
PM
4201}
4202
2b5ec1a5
YY
4203static inline void ipvs_reset(struct sk_buff *skb)
4204{
4205#if IS_ENABLED(CONFIG_IP_VS)
4206 skb->ipvs_property = 0;
4207#endif
4208}
4209
de8bda1d 4210/* Note: This doesn't put any conntrack info in dst. */
b1937227
ED
4211static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4212 bool copy)
edda553c 4213{
5f79e0f9 4214#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc
FW
4215 dst->_nfct = src->_nfct;
4216 nf_conntrack_get(skb_nfct(src));
2fc72c7b 4217#endif
478b360a 4218#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
b1937227
ED
4219 if (copy)
4220 dst->nf_trace = src->nf_trace;
478b360a 4221#endif
edda553c
YK
4222}
4223
e7ac05f3
YK
4224static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4225{
e7ac05f3 4226#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
a9e419dc 4227 nf_conntrack_put(skb_nfct(dst));
e7ac05f3 4228#endif
b1937227 4229 __nf_copy(dst, src, true);
e7ac05f3
YK
4230}
4231
984bc16c
JM
4232#ifdef CONFIG_NETWORK_SECMARK
4233static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4234{
4235 to->secmark = from->secmark;
4236}
4237
4238static inline void skb_init_secmark(struct sk_buff *skb)
4239{
4240 skb->secmark = 0;
4241}
4242#else
4243static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4244{ }
4245
4246static inline void skb_init_secmark(struct sk_buff *skb)
4247{ }
4248#endif
4249
7af8f4ca
FW
4250static inline int secpath_exists(const struct sk_buff *skb)
4251{
4252#ifdef CONFIG_XFRM
4165079b 4253 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
7af8f4ca
FW
4254#else
4255 return 0;
4256#endif
4257}
4258
574f7194
EB
4259static inline bool skb_irq_freeable(const struct sk_buff *skb)
4260{
4261 return !skb->destructor &&
7af8f4ca 4262 !secpath_exists(skb) &&
cb9c6836 4263 !skb_nfct(skb) &&
574f7194
EB
4264 !skb->_skb_refdst &&
4265 !skb_has_frag_list(skb);
4266}
4267
f25f4e44
PWJ
4268static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4269{
f25f4e44 4270 skb->queue_mapping = queue_mapping;
f25f4e44
PWJ
4271}
4272
9247744e 4273static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
4e3ab47a 4274{
4e3ab47a 4275 return skb->queue_mapping;
4e3ab47a
PE
4276}
4277
f25f4e44
PWJ
4278static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4279{
f25f4e44 4280 to->queue_mapping = from->queue_mapping;
f25f4e44
PWJ
4281}
4282
d5a9e24a
DM
4283static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4284{
4285 skb->queue_mapping = rx_queue + 1;
4286}
4287
9247744e 4288static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
d5a9e24a
DM
4289{
4290 return skb->queue_mapping - 1;
4291}
4292
9247744e 4293static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
d5a9e24a 4294{
a02cec21 4295 return skb->queue_mapping != 0;
d5a9e24a
DM
4296}
4297
4ff06203
JA
4298static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4299{
4300 skb->dst_pending_confirm = val;
4301}
4302
4303static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4304{
4305 return skb->dst_pending_confirm != 0;
4306}
4307
2294be0f 4308static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
def8b4fa 4309{
0b3d8e08 4310#ifdef CONFIG_XFRM
4165079b 4311 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
def8b4fa 4312#else
def8b4fa 4313 return NULL;
def8b4fa 4314#endif
0b3d8e08 4315}
def8b4fa 4316
68c33163
PS
4317/* Keeps track of mac header offset relative to skb->head.
4318 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4319 * For non-tunnel skb it points to skb_mac_header() and for
3347c960
ED
4320 * tunnel skb it points to outer mac header.
4321 * Keeps track of level of encapsulation of network headers.
4322 */
68c33163 4323struct skb_gso_cb {
802ab55a
AD
4324 union {
4325 int mac_offset;
4326 int data_offset;
4327 };
3347c960 4328 int encap_level;
76443456 4329 __wsum csum;
7e2b10c1 4330 __u16 csum_start;
68c33163 4331};
9207f9d4
KK
4332#define SKB_SGO_CB_OFFSET 32
4333#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
68c33163
PS
4334
4335static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4336{
4337 return (skb_mac_header(inner_skb) - inner_skb->head) -
4338 SKB_GSO_CB(inner_skb)->mac_offset;
4339}
4340
1e2bd517
PS
4341static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4342{
4343 int new_headroom, headroom;
4344 int ret;
4345
4346 headroom = skb_headroom(skb);
4347 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4348 if (ret)
4349 return ret;
4350
4351 new_headroom = skb_headroom(skb);
4352 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4353 return 0;
4354}
4355
08b64fcc
AD
4356static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4357{
4358 /* Do not update partial checksums if remote checksum is enabled. */
4359 if (skb->remcsum_offload)
4360 return;
4361
4362 SKB_GSO_CB(skb)->csum = res;
4363 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4364}
4365
7e2b10c1
TH
4366/* Compute the checksum for a gso segment. First compute the checksum value
4367 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4368 * then add in skb->csum (checksum from csum_start to end of packet).
4369 * skb->csum and csum_start are then updated to reflect the checksum of the
4370 * resultant packet starting from the transport header-- the resultant checksum
4371 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4372 * header.
4373 */
4374static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4375{
76443456
AD
4376 unsigned char *csum_start = skb_transport_header(skb);
4377 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4378 __wsum partial = SKB_GSO_CB(skb)->csum;
7e2b10c1 4379
76443456
AD
4380 SKB_GSO_CB(skb)->csum = res;
4381 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
7e2b10c1 4382
76443456 4383 return csum_fold(csum_partial(csum_start, plen, partial));
7e2b10c1
TH
4384}
4385
bdcc0924 4386static inline bool skb_is_gso(const struct sk_buff *skb)
89114afd
HX
4387{
4388 return skb_shinfo(skb)->gso_size;
4389}
4390
36a8f39e 4391/* Note: Should be called only if skb_is_gso(skb) is true */
bdcc0924 4392static inline bool skb_is_gso_v6(const struct sk_buff *skb)
eabd7e35
BG
4393{
4394 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4395}
4396
d02f51cb
DA
4397/* Note: Should be called only if skb_is_gso(skb) is true */
4398static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4399{
4400 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4401}
4402
4c3024de 4403/* Note: Should be called only if skb_is_gso(skb) is true */
b90efd22
WB
4404static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4405{
4c3024de 4406 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
b90efd22
WB
4407}
4408
5293efe6
DB
4409static inline void skb_gso_reset(struct sk_buff *skb)
4410{
4411 skb_shinfo(skb)->gso_size = 0;
4412 skb_shinfo(skb)->gso_segs = 0;
4413 skb_shinfo(skb)->gso_type = 0;
4414}
4415
d02f51cb
DA
4416static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4417 u16 increment)
4418{
4419 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4420 return;
4421 shinfo->gso_size += increment;
4422}
4423
4424static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4425 u16 decrement)
4426{
4427 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4428 return;
4429 shinfo->gso_size -= decrement;
4430}
4431
7965bd4d 4432void __skb_warn_lro_forwarding(const struct sk_buff *skb);
4497b076
BH
4433
4434static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4435{
4436 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4437 * wanted then gso_type will be set. */
05bdd2f1
ED
4438 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4439
b78462eb
AD
4440 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4441 unlikely(shinfo->gso_type == 0)) {
4497b076
BH
4442 __skb_warn_lro_forwarding(skb);
4443 return true;
4444 }
4445 return false;
4446}
4447
35fc92a9
HX
4448static inline void skb_forward_csum(struct sk_buff *skb)
4449{
4450 /* Unfortunately we don't support this one. Any brave souls? */
4451 if (skb->ip_summed == CHECKSUM_COMPLETE)
4452 skb->ip_summed = CHECKSUM_NONE;
4453}
4454
bc8acf2c
ED
4455/**
4456 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4457 * @skb: skb to check
4458 *
4459 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4460 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4461 * use this helper, to document places where we make this assertion.
4462 */
05bdd2f1 4463static inline void skb_checksum_none_assert(const struct sk_buff *skb)
bc8acf2c
ED
4464{
4465#ifdef DEBUG
4466 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4467#endif
4468}
4469
f35d9d8a 4470bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
a6686f2f 4471
ed1f50c3 4472int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
9afd85c9
LL
4473struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4474 unsigned int transport_len,
4475 __sum16(*skb_chkf)(struct sk_buff *skb));
ed1f50c3 4476
3a7c1ee4
AD
4477/**
4478 * skb_head_is_locked - Determine if the skb->head is locked down
4479 * @skb: skb to check
4480 *
4481 * The head on skbs build around a head frag can be removed if they are
4482 * not cloned. This function returns true if the skb head is locked down
4483 * due to either being allocated via kmalloc, or by being a clone with
4484 * multiple references to the head.
4485 */
4486static inline bool skb_head_is_locked(const struct sk_buff *skb)
4487{
4488 return !skb->head_frag || skb_cloned(skb);
4489}
fe6cc55f 4490
179bc67f
EC
4491/* Local Checksum Offload.
4492 * Compute outer checksum based on the assumption that the
4493 * inner checksum will be offloaded later.
d0dcde64 4494 * See Documentation/networking/checksum-offloads.rst for
e8ae7b00 4495 * explanation of how this works.
179bc67f
EC
4496 * Fill in outer checksum adjustment (e.g. with sum of outer
4497 * pseudo-header) before calling.
4498 * Also ensure that inner checksum is in linear data area.
4499 */
4500static inline __wsum lco_csum(struct sk_buff *skb)
4501{
9e74a6da
AD
4502 unsigned char *csum_start = skb_checksum_start(skb);
4503 unsigned char *l4_hdr = skb_transport_header(skb);
4504 __wsum partial;
179bc67f
EC
4505
4506 /* Start with complement of inner checksum adjustment */
9e74a6da
AD
4507 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4508 skb->csum_offset));
4509
179bc67f 4510 /* Add in checksum of our headers (incl. outer checksum
9e74a6da 4511 * adjustment filled in by caller) and return result.
179bc67f 4512 */
9e74a6da 4513 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
179bc67f
EC
4514}
4515
1da177e4
LT
4516#endif /* __KERNEL__ */
4517#endif /* _LINUX_SKBUFF_H */