Commit | Line | Data |
---|---|---|
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 <net/checksum.h> |
a80958f4 | 27 | #include <linux/rcupdate.h> |
131ea667 | 28 | #include <linux/dma-mapping.h> |
c8f44aff | 29 | #include <linux/netdev_features.h> |
1bd758eb | 30 | #include <net/flow_dissector.h> |
72b31f72 | 31 | #include <linux/in6.h> |
8b10cab6 | 32 | #include <linux/if_packet.h> |
f35f8219 | 33 | #include <linux/llist.h> |
f70ea018 | 34 | #include <net/flow.h> |
261db6c2 JS |
35 | #if IS_ENABLED(CONFIG_NF_CONNTRACK) |
36 | #include <linux/netfilter/nf_conntrack_common.h> | |
37 | #endif | |
66e4c8d9 | 38 | #include <net/net_debug.h> |
5b8285cc | 39 | #include <net/dropreason-core.h> |
21d2e673 | 40 | #include <net/netmem.h> |
1da177e4 | 41 | |
9facd941 JK |
42 | /** |
43 | * DOC: skb checksums | |
44 | * | |
45 | * The interface for checksum offload between the stack and networking drivers | |
7a6ae71b TH |
46 | * is as follows... |
47 | * | |
9facd941 JK |
48 | * IP checksum related features |
49 | * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~ | |
7a6ae71b TH |
50 | * |
51 | * Drivers advertise checksum offload capabilities in the features of a device. | |
db1f00fb DC |
52 | * From the stack's point of view these are capabilities offered by the driver. |
53 | * A driver typically only advertises features that it is capable of offloading | |
7a6ae71b TH |
54 | * to its device. |
55 | * | |
9facd941 JK |
56 | * .. flat-table:: Checksum related device features |
57 | * :widths: 1 10 | |
58 | * | |
59 | * * - %NETIF_F_HW_CSUM | |
60 | * - The driver (or its device) is able to compute one | |
61 | * IP (one's complement) checksum for any combination | |
62 | * of protocols or protocol layering. The checksum is | |
63 | * computed and set in a packet per the CHECKSUM_PARTIAL | |
64 | * interface (see below). | |
65 | * | |
66 | * * - %NETIF_F_IP_CSUM | |
67 | * - Driver (device) is only able to checksum plain | |
68 | * TCP or UDP packets over IPv4. These are specifically | |
69 | * unencapsulated packets of the form IPv4|TCP or | |
70 | * IPv4|UDP where the Protocol field in the IPv4 header | |
71 | * is TCP or UDP. The IPv4 header may contain IP options. | |
72 | * This feature cannot be set in features for a device | |
73 | * with NETIF_F_HW_CSUM also set. This feature is being | |
74 | * DEPRECATED (see below). | |
75 | * | |
76 | * * - %NETIF_F_IPV6_CSUM | |
77 | * - Driver (device) is only able to checksum plain | |
78 | * TCP or UDP packets over IPv6. These are specifically | |
79 | * unencapsulated packets of the form IPv6|TCP or | |
80 | * IPv6|UDP where the Next Header field in the IPv6 | |
81 | * header is either TCP or UDP. IPv6 extension headers | |
82 | * are not supported with this feature. This feature | |
83 | * cannot be set in features for a device with | |
84 | * NETIF_F_HW_CSUM also set. This feature is being | |
85 | * DEPRECATED (see below). | |
86 | * | |
87 | * * - %NETIF_F_RXCSUM | |
88 | * - Driver (device) performs receive checksum offload. | |
89 | * This flag is only used to disable the RX checksum | |
90 | * feature for a device. The stack will accept receive | |
91 | * checksum indication in packets received on a device | |
92 | * regardless of whether NETIF_F_RXCSUM is set. | |
93 | * | |
94 | * Checksumming of received packets by device | |
95 | * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ | |
96 | * | |
97 | * Indication of checksum verification is set in &sk_buff.ip_summed. | |
98 | * Possible values are: | |
99 | * | |
100 | * - %CHECKSUM_NONE | |
78ea85f1 | 101 | * |
7a6ae71b | 102 | * Device did not checksum this packet e.g. due to lack of capabilities. |
78ea85f1 DB |
103 | * The packet contains full (though not verified) checksum in packet but |
104 | * not in skb->csum. Thus, skb->csum is undefined in this case. | |
105 | * | |
9facd941 | 106 | * - %CHECKSUM_UNNECESSARY |
78ea85f1 DB |
107 | * |
108 | * The hardware you're dealing with doesn't calculate the full checksum | |
9facd941 JK |
109 | * (as in %CHECKSUM_COMPLETE), but it does parse headers and verify checksums |
110 | * for specific protocols. For such packets it will set %CHECKSUM_UNNECESSARY | |
111 | * if their checksums are okay. &sk_buff.csum is still undefined in this case | |
7a6ae71b TH |
112 | * though. A driver or device must never modify the checksum field in the |
113 | * packet even if checksum is verified. | |
77cffe23 | 114 | * |
9facd941 JK |
115 | * %CHECKSUM_UNNECESSARY is applicable to following protocols: |
116 | * | |
117 | * - TCP: IPv6 and IPv4. | |
118 | * - UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a | |
77cffe23 TH |
119 | * zero UDP checksum for either IPv4 or IPv6, the networking stack |
120 | * may perform further validation in this case. | |
9facd941 JK |
121 | * - GRE: only if the checksum is present in the header. |
122 | * - SCTP: indicates the CRC in SCTP header has been validated. | |
123 | * - FCOE: indicates the CRC in FC frame has been validated. | |
77cffe23 | 124 | * |
9facd941 JK |
125 | * &sk_buff.csum_level indicates the number of consecutive checksums found in |
126 | * the packet minus one that have been verified as %CHECKSUM_UNNECESSARY. | |
77cffe23 TH |
127 | * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet |
128 | * and a device is able to verify the checksums for UDP (possibly zero), | |
9facd941 | 129 | * GRE (checksum flag is set) and TCP, &sk_buff.csum_level would be set to |
77cffe23 | 130 | * two. If the device were only able to verify the UDP checksum and not |
db1f00fb | 131 | * GRE, either because it doesn't support GRE checksum or because GRE |
77cffe23 TH |
132 | * checksum is bad, skb->csum_level would be set to zero (TCP checksum is |
133 | * not considered in this case). | |
78ea85f1 | 134 | * |
9facd941 | 135 | * - %CHECKSUM_COMPLETE |
78ea85f1 DB |
136 | * |
137 | * This is the most generic way. The device supplied checksum of the _whole_ | |
9facd941 | 138 | * packet as seen by netif_rx() and fills in &sk_buff.csum. This means the |
78ea85f1 DB |
139 | * hardware doesn't need to parse L3/L4 headers to implement this. |
140 | * | |
b4759dcd | 141 | * Notes: |
9facd941 | 142 | * |
b4759dcd DC |
143 | * - Even if device supports only some protocols, but is able to produce |
144 | * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY. | |
145 | * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols. | |
78ea85f1 | 146 | * |
9facd941 | 147 | * - %CHECKSUM_PARTIAL |
78ea85f1 | 148 | * |
6edec0e6 TH |
149 | * A checksum is set up to be offloaded to a device as described in the |
150 | * output description for CHECKSUM_PARTIAL. This may occur on a packet | |
78ea85f1 | 151 | * received directly from another Linux OS, e.g., a virtualized Linux kernel |
6edec0e6 TH |
152 | * on the same host, or it may be set in the input path in GRO or remote |
153 | * checksum offload. For the purposes of checksum verification, the checksum | |
154 | * referred to by skb->csum_start + skb->csum_offset and any preceding | |
155 | * checksums in the packet are considered verified. Any checksums in the | |
156 | * packet that are after the checksum being offloaded are not considered to | |
157 | * be verified. | |
78ea85f1 | 158 | * |
9facd941 JK |
159 | * Checksumming on transmit for non-GSO |
160 | * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ | |
78ea85f1 | 161 | * |
9facd941 JK |
162 | * The stack requests checksum offload in the &sk_buff.ip_summed for a packet. |
163 | * Values are: | |
164 | * | |
165 | * - %CHECKSUM_PARTIAL | |
78ea85f1 | 166 | * |
7a6ae71b | 167 | * The driver is required to checksum the packet as seen by hard_start_xmit() |
9facd941 JK |
168 | * from &sk_buff.csum_start up to the end, and to record/write the checksum at |
169 | * offset &sk_buff.csum_start + &sk_buff.csum_offset. | |
170 | * A driver may verify that the | |
7a6ae71b | 171 | * csum_start and csum_offset values are valid values given the length and |
db1f00fb DC |
172 | * offset of the packet, but it should not attempt to validate that the |
173 | * checksum refers to a legitimate transport layer checksum -- it is the | |
7a6ae71b TH |
174 | * purview of the stack to validate that csum_start and csum_offset are set |
175 | * correctly. | |
176 | * | |
177 | * When the stack requests checksum offload for a packet, the driver MUST | |
178 | * ensure that the checksum is set correctly. A driver can either offload the | |
179 | * checksum calculation to the device, or call skb_checksum_help (in the case | |
180 | * that the device does not support offload for a particular checksum). | |
181 | * | |
9facd941 JK |
182 | * %NETIF_F_IP_CSUM and %NETIF_F_IPV6_CSUM are being deprecated in favor of |
183 | * %NETIF_F_HW_CSUM. New devices should use %NETIF_F_HW_CSUM to indicate | |
43c26a1a | 184 | * checksum offload capability. |
9facd941 | 185 | * skb_csum_hwoffload_help() can be called to resolve %CHECKSUM_PARTIAL based |
43c26a1a | 186 | * on network device checksumming capabilities: if a packet does not match |
9facd941 JK |
187 | * them, skb_checksum_help() or skb_crc32c_help() (depending on the value of |
188 | * &sk_buff.csum_not_inet, see :ref:`crc`) | |
189 | * is called to resolve the checksum. | |
78ea85f1 | 190 | * |
9facd941 | 191 | * - %CHECKSUM_NONE |
78ea85f1 | 192 | * |
7a6ae71b TH |
193 | * The skb was already checksummed by the protocol, or a checksum is not |
194 | * required. | |
78ea85f1 | 195 | * |
9facd941 | 196 | * - %CHECKSUM_UNNECESSARY |
78ea85f1 | 197 | * |
db1f00fb | 198 | * This has the same meaning as CHECKSUM_NONE for checksum offload on |
7a6ae71b | 199 | * output. |
78ea85f1 | 200 | * |
9facd941 JK |
201 | * - %CHECKSUM_COMPLETE |
202 | * | |
7a6ae71b | 203 | * Not used in checksum output. If a driver observes a packet with this value |
9facd941 JK |
204 | * set in skbuff, it should treat the packet as if %CHECKSUM_NONE were set. |
205 | * | |
206 | * .. _crc: | |
207 | * | |
208 | * Non-IP checksum (CRC) offloads | |
209 | * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ | |
210 | * | |
211 | * .. flat-table:: | |
212 | * :widths: 1 10 | |
213 | * | |
214 | * * - %NETIF_F_SCTP_CRC | |
215 | * - This feature indicates that a device is capable of | |
216 | * offloading the SCTP CRC in a packet. To perform this offload the stack | |
217 | * will set csum_start and csum_offset accordingly, set ip_summed to | |
218 | * %CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication | |
219 | * in the skbuff that the %CHECKSUM_PARTIAL refers to CRC32c. | |
220 | * A driver that supports both IP checksum offload and SCTP CRC32c offload | |
221 | * must verify which offload is configured for a packet by testing the | |
222 | * value of &sk_buff.csum_not_inet; skb_crc32c_csum_help() is provided to | |
223 | * resolve %CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1. | |
224 | * | |
225 | * * - %NETIF_F_FCOE_CRC | |
226 | * - This feature indicates that a device is capable of offloading the FCOE | |
227 | * CRC in a packet. To perform this offload the stack will set ip_summed | |
228 | * to %CHECKSUM_PARTIAL and set csum_start and csum_offset | |
229 | * accordingly. Note that there is no indication in the skbuff that the | |
230 | * %CHECKSUM_PARTIAL refers to an FCOE checksum, so a driver that supports | |
231 | * both IP checksum offload and FCOE CRC offload must verify which offload | |
232 | * is configured for a packet, presumably by inspecting packet headers. | |
233 | * | |
234 | * Checksumming on output with GSO | |
235 | * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ | |
236 | * | |
237 | * In the case of a GSO packet (skb_is_gso() is true), checksum offload | |
7a6ae71b | 238 | * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the |
9facd941 | 239 | * gso_type is %SKB_GSO_TCPV4 or %SKB_GSO_TCPV6, TCP checksum offload as |
7a6ae71b | 240 | * part of the GSO operation is implied. If a checksum is being offloaded |
9facd941 | 241 | * with GSO then ip_summed is %CHECKSUM_PARTIAL, and both csum_start and |
db1f00fb DC |
242 | * csum_offset are set to refer to the outermost checksum being offloaded |
243 | * (two offloaded checksums are possible with UDP encapsulation). | |
78ea85f1 DB |
244 | */ |
245 | ||
60476372 | 246 | /* Don't change this without changing skb_csum_unnecessary! */ |
78ea85f1 DB |
247 | #define CHECKSUM_NONE 0 |
248 | #define CHECKSUM_UNNECESSARY 1 | |
249 | #define CHECKSUM_COMPLETE 2 | |
250 | #define CHECKSUM_PARTIAL 3 | |
1da177e4 | 251 | |
77cffe23 TH |
252 | /* Maximum value in skb->csum_level */ |
253 | #define SKB_MAX_CSUM_LEVEL 3 | |
254 | ||
0bec8c88 | 255 | #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES) |
fc910a27 | 256 | #define SKB_WITH_OVERHEAD(X) \ |
deea84b0 | 257 | ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) |
115f1a5c ED |
258 | |
259 | /* For X bytes available in skb->head, what is the minimal | |
260 | * allocation needed, knowing struct skb_shared_info needs | |
261 | * to be aligned. | |
262 | */ | |
263 | #define SKB_HEAD_ALIGN(X) (SKB_DATA_ALIGN(X) + \ | |
264 | SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) | |
265 | ||
fc910a27 DM |
266 | #define SKB_MAX_ORDER(X, ORDER) \ |
267 | SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X)) | |
1da177e4 LT |
268 | #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0)) |
269 | #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2)) | |
270 | ||
87fb4b7b ED |
271 | /* return minimum truesize of one skb containing X bytes of data */ |
272 | #define SKB_TRUESIZE(X) ((X) + \ | |
273 | SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \ | |
274 | SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) | |
275 | ||
7999096f | 276 | struct ahash_request; |
1da177e4 | 277 | struct net_device; |
716ea3a7 | 278 | struct scatterlist; |
9c55e01c | 279 | struct pipe_inode_info; |
a8f820aa | 280 | struct iov_iter; |
fd11a83d | 281 | struct napi_struct; |
d58e468b PP |
282 | struct bpf_prog; |
283 | union bpf_attr; | |
df5042f4 | 284 | struct skb_ext; |
2195e2a0 | 285 | struct ts_config; |
1da177e4 | 286 | |
34666d46 | 287 | #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) |
1da177e4 | 288 | struct nf_bridge_info { |
3eaf4025 FW |
289 | enum { |
290 | BRNF_PROTO_UNCHANGED, | |
291 | BRNF_PROTO_8021Q, | |
292 | BRNF_PROTO_PPPOE | |
7fb48c5b | 293 | } orig_proto:8; |
72b1e5e4 FW |
294 | u8 pkt_otherhost:1; |
295 | u8 in_prerouting:1; | |
296 | u8 bridged_dnat:1; | |
94623f57 | 297 | u8 sabotage_in_done:1; |
411ffb4f | 298 | __u16 frag_max_size; |
98748088 | 299 | int physinif; |
63cdbc06 FW |
300 | |
301 | /* always valid & non-NULL from FORWARD on, for physdev match */ | |
302 | struct net_device *physoutdev; | |
7fb48c5b | 303 | union { |
72b1e5e4 | 304 | /* prerouting: detect dnat in orig/reply direction */ |
72b31f72 BT |
305 | __be32 ipv4_daddr; |
306 | struct in6_addr ipv6_daddr; | |
72b1e5e4 FW |
307 | |
308 | /* after prerouting + nat detected: store original source | |
309 | * mac since neigh resolution overwrites it, only used while | |
310 | * skb is out in neigh layer. | |
311 | */ | |
312 | char neigh_header[8]; | |
72b31f72 | 313 | }; |
1da177e4 LT |
314 | }; |
315 | #endif | |
316 | ||
95a7233c PB |
317 | #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT) |
318 | /* Chain in tc_skb_ext will be used to share the tc chain with | |
319 | * ovs recirc_id. It will be set to the current chain by tc | |
320 | * and read by ovs to recirc_id. | |
321 | */ | |
322 | struct tc_skb_ext { | |
80cd22c3 PB |
323 | union { |
324 | u64 act_miss_cookie; | |
325 | __u32 chain; | |
326 | }; | |
038ebb1a | 327 | __u16 mru; |
635d448a | 328 | __u16 zone; |
6f022c2d PB |
329 | u8 post_ct:1; |
330 | u8 post_ct_snat:1; | |
331 | u8 post_ct_dnat:1; | |
80cd22c3 | 332 | u8 act_miss:1; /* Set if act_miss_cookie is used */ |
7b4858df | 333 | u8 l2_miss:1; /* Set by bridge upon FDB or MDB miss */ |
95a7233c PB |
334 | }; |
335 | #endif | |
336 | ||
1da177e4 | 337 | struct sk_buff_head { |
1a2fb220 KC |
338 | /* These two members must be first to match sk_buff. */ |
339 | struct_group_tagged(sk_buff_list, list, | |
340 | struct sk_buff *next; | |
341 | struct sk_buff *prev; | |
342 | ); | |
1da177e4 LT |
343 | |
344 | __u32 qlen; | |
345 | spinlock_t lock; | |
346 | }; | |
347 | ||
348 | struct sk_buff; | |
349 | ||
3948b059 ED |
350 | #ifndef CONFIG_MAX_SKB_FRAGS |
351 | # define CONFIG_MAX_SKB_FRAGS 17 | |
a715dea3 | 352 | #endif |
3948b059 ED |
353 | |
354 | #define MAX_SKB_FRAGS CONFIG_MAX_SKB_FRAGS | |
355 | ||
3953c46c MRL |
356 | /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to |
357 | * segment using its current segmentation instead. | |
358 | */ | |
359 | #define GSO_BY_FRAGS 0xFFFF | |
360 | ||
21d2e673 MA |
361 | typedef struct skb_frag { |
362 | netmem_ref netmem; | |
363 | unsigned int len; | |
364 | unsigned int offset; | |
365 | } skb_frag_t; | |
1da177e4 | 366 | |
161e6137 | 367 | /** |
7240b60c | 368 | * skb_frag_size() - Returns the size of a skb fragment |
161e6137 PT |
369 | * @frag: skb fragment |
370 | */ | |
9e903e08 ED |
371 | static inline unsigned int skb_frag_size(const skb_frag_t *frag) |
372 | { | |
21d2e673 | 373 | return frag->len; |
9e903e08 ED |
374 | } |
375 | ||
161e6137 | 376 | /** |
7240b60c | 377 | * skb_frag_size_set() - Sets the size of a skb fragment |
161e6137 PT |
378 | * @frag: skb fragment |
379 | * @size: size of fragment | |
380 | */ | |
9e903e08 ED |
381 | static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size) |
382 | { | |
21d2e673 | 383 | frag->len = size; |
9e903e08 ED |
384 | } |
385 | ||
161e6137 | 386 | /** |
7240b60c | 387 | * skb_frag_size_add() - Increments the size of a skb fragment by @delta |
161e6137 PT |
388 | * @frag: skb fragment |
389 | * @delta: value to add | |
390 | */ | |
9e903e08 ED |
391 | static inline void skb_frag_size_add(skb_frag_t *frag, int delta) |
392 | { | |
21d2e673 | 393 | frag->len += delta; |
9e903e08 ED |
394 | } |
395 | ||
161e6137 | 396 | /** |
7240b60c | 397 | * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta |
161e6137 PT |
398 | * @frag: skb fragment |
399 | * @delta: value to subtract | |
400 | */ | |
9e903e08 ED |
401 | static inline void skb_frag_size_sub(skb_frag_t *frag, int delta) |
402 | { | |
21d2e673 | 403 | frag->len -= delta; |
9e903e08 ED |
404 | } |
405 | ||
161e6137 PT |
406 | /** |
407 | * skb_frag_must_loop - Test if %p is a high memory page | |
408 | * @p: fragment's page | |
409 | */ | |
c613c209 WB |
410 | static inline bool skb_frag_must_loop(struct page *p) |
411 | { | |
412 | #if defined(CONFIG_HIGHMEM) | |
29766bcf | 413 | if (IS_ENABLED(CONFIG_DEBUG_KMAP_LOCAL_FORCE_MAP) || PageHighMem(p)) |
c613c209 WB |
414 | return true; |
415 | #endif | |
416 | return false; | |
417 | } | |
418 | ||
419 | /** | |
420 | * skb_frag_foreach_page - loop over pages in a fragment | |
421 | * | |
422 | * @f: skb frag to operate on | |
21d2e673 | 423 | * @f_off: offset from start of f->netmem |
c613c209 WB |
424 | * @f_len: length from f_off to loop over |
425 | * @p: (temp var) current page | |
426 | * @p_off: (temp var) offset from start of current page, | |
427 | * non-zero only on first page. | |
428 | * @p_len: (temp var) length in current page, | |
429 | * < PAGE_SIZE only on first and last page. | |
430 | * @copied: (temp var) length so far, excluding current p_len. | |
431 | * | |
432 | * A fragment can hold a compound page, in which case per-page | |
433 | * operations, notably kmap_atomic, must be called for each | |
434 | * regular page. | |
435 | */ | |
436 | #define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \ | |
437 | for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \ | |
438 | p_off = (f_off) & (PAGE_SIZE - 1), \ | |
439 | p_len = skb_frag_must_loop(p) ? \ | |
440 | min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \ | |
441 | copied = 0; \ | |
442 | copied < f_len; \ | |
443 | copied += p_len, p++, p_off = 0, \ | |
444 | p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \ | |
445 | ||
ac45f602 | 446 | /** |
d3a21be8 | 447 | * struct skb_shared_hwtstamps - hardware time stamps |
97dc7cd9 GE |
448 | * @hwtstamp: hardware time stamp transformed into duration |
449 | * since arbitrary point in time | |
450 | * @netdev_data: address/cookie of network device driver used as | |
451 | * reference to actual hardware time stamp | |
ac45f602 PO |
452 | * |
453 | * Software time stamps generated by ktime_get_real() are stored in | |
4d276eb6 | 454 | * skb->tstamp. |
ac45f602 PO |
455 | * |
456 | * hwtstamps can only be compared against other hwtstamps from | |
457 | * the same device. | |
458 | * | |
459 | * This structure is attached to packets as part of the | |
460 | * &skb_shared_info. Use skb_hwtstamps() to get a pointer. | |
461 | */ | |
462 | struct skb_shared_hwtstamps { | |
97dc7cd9 GE |
463 | union { |
464 | ktime_t hwtstamp; | |
465 | void *netdev_data; | |
466 | }; | |
ac45f602 PO |
467 | }; |
468 | ||
2244d07b OH |
469 | /* Definitions for tx_flags in struct skb_shared_info */ |
470 | enum { | |
471 | /* generate hardware time stamp */ | |
472 | SKBTX_HW_TSTAMP = 1 << 0, | |
473 | ||
e7fd2885 | 474 | /* generate software time stamp when queueing packet to NIC */ |
2244d07b OH |
475 | SKBTX_SW_TSTAMP = 1 << 1, |
476 | ||
477 | /* device driver is going to provide hardware time stamp */ | |
478 | SKBTX_IN_PROGRESS = 1 << 2, | |
479 | ||
51eb7492 GE |
480 | /* generate hardware time stamp based on cycles if supported */ |
481 | SKBTX_HW_TSTAMP_USE_CYCLES = 1 << 3, | |
482 | ||
6e3e939f | 483 | /* generate wifi status information (where possible) */ |
62b1a8ab | 484 | SKBTX_WIFI_STATUS = 1 << 4, |
c9af6db4 | 485 | |
97dc7cd9 GE |
486 | /* determine hardware time stamp based on time or cycles */ |
487 | SKBTX_HW_TSTAMP_NETDEV = 1 << 5, | |
488 | ||
e7fd2885 WB |
489 | /* generate software time stamp when entering packet scheduling */ |
490 | SKBTX_SCHED_TSTAMP = 1 << 6, | |
a6686f2f SM |
491 | }; |
492 | ||
e1c8a607 | 493 | #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \ |
0a2cf20c | 494 | SKBTX_SCHED_TSTAMP) |
51eb7492 GE |
495 | #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | \ |
496 | SKBTX_HW_TSTAMP_USE_CYCLES | \ | |
497 | SKBTX_ANY_SW_TSTAMP) | |
f24b9be5 | 498 | |
06b4feb3 JL |
499 | /* Definitions for flags in struct skb_shared_info */ |
500 | enum { | |
501 | /* use zcopy routines */ | |
502 | SKBFL_ZEROCOPY_ENABLE = BIT(0), | |
503 | ||
504 | /* This indicates at least one fragment might be overwritten | |
505 | * (as in vmsplice(), sendfile() ...) | |
506 | * If we need to compute a TX checksum, we'll need to copy | |
507 | * all frags to avoid possible bad checksum | |
508 | */ | |
509 | SKBFL_SHARED_FRAG = BIT(1), | |
9b65b17d TA |
510 | |
511 | /* segment contains only zerocopy data and should not be | |
512 | * charged to the kernel memory. | |
513 | */ | |
514 | SKBFL_PURE_ZEROCOPY = BIT(2), | |
2e07a521 PB |
515 | |
516 | SKBFL_DONT_ORPHAN = BIT(3), | |
753f1ca4 PB |
517 | |
518 | /* page references are managed by the ubuf_info, so it's safe to | |
519 | * use frags only up until ubuf_info is released | |
520 | */ | |
521 | SKBFL_MANAGED_FRAG_REFS = BIT(4), | |
06b4feb3 JL |
522 | }; |
523 | ||
524 | #define SKBFL_ZEROCOPY_FRAG (SKBFL_ZEROCOPY_ENABLE | SKBFL_SHARED_FRAG) | |
2e07a521 | 525 | #define SKBFL_ALL_ZEROCOPY (SKBFL_ZEROCOPY_FRAG | SKBFL_PURE_ZEROCOPY | \ |
753f1ca4 | 526 | SKBFL_DONT_ORPHAN | SKBFL_MANAGED_FRAG_REFS) |
06b4feb3 | 527 | |
7ab4f16f PB |
528 | struct ubuf_info_ops { |
529 | void (*complete)(struct sk_buff *, struct ubuf_info *, | |
530 | bool zerocopy_success); | |
65bada80 PB |
531 | /* has to be compatible with skb_zcopy_set() */ |
532 | int (*link_skb)(struct sk_buff *skb, struct ubuf_info *uarg); | |
7ab4f16f PB |
533 | }; |
534 | ||
a6686f2f SM |
535 | /* |
536 | * The callback notifies userspace to release buffers when skb DMA is done in | |
537 | * lower device, the skb last reference should be 0 when calling this. | |
e19d6763 MT |
538 | * The zerocopy_success argument is true if zero copy transmit occurred, |
539 | * false on data copy or out of memory error caused by data copy attempt. | |
ca8f4fb2 MT |
540 | * The ctx field is used to track device context. |
541 | * The desc field is used to track userspace buffer index. | |
a6686f2f SM |
542 | */ |
543 | struct ubuf_info { | |
7ab4f16f | 544 | const struct ubuf_info_ops *ops; |
c1d1b437 | 545 | refcount_t refcnt; |
04c2d33e | 546 | u8 flags; |
ac45f602 PO |
547 | }; |
548 | ||
6eaab4df PB |
549 | struct ubuf_info_msgzc { |
550 | struct ubuf_info ubuf; | |
551 | ||
552 | union { | |
553 | struct { | |
554 | unsigned long desc; | |
555 | void *ctx; | |
556 | }; | |
557 | struct { | |
558 | u32 id; | |
559 | u16 len; | |
560 | u16 zerocopy:1; | |
561 | u32 bytelen; | |
562 | }; | |
563 | }; | |
564 | ||
e7d2b510 PB |
565 | struct mmpin { |
566 | struct user_struct *user; | |
567 | unsigned int num_pg; | |
568 | } mmp; | |
6eaab4df PB |
569 | }; |
570 | ||
52267790 | 571 | #define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg)) |
6eaab4df PB |
572 | #define uarg_to_msgzc(ubuf_ptr) container_of((ubuf_ptr), struct ubuf_info_msgzc, \ |
573 | ubuf) | |
52267790 | 574 | |
6f89dbce SV |
575 | int mm_account_pinned_pages(struct mmpin *mmp, size_t size); |
576 | void mm_unaccount_pinned_pages(struct mmpin *mmp); | |
577 | ||
48eb03dd SF |
578 | /* Preserve some data across TX submission and completion. |
579 | * | |
580 | * Note, this state is stored in the driver. Extending the layout | |
581 | * might need some special care. | |
582 | */ | |
583 | struct xsk_tx_metadata_compl { | |
584 | __u64 *tx_timestamp; | |
585 | }; | |
586 | ||
1da177e4 LT |
587 | /* This data is invariant across clones and lives at |
588 | * the end of the header data, ie. at skb->end. | |
589 | */ | |
590 | struct skb_shared_info { | |
06b4feb3 | 591 | __u8 flags; |
de8f3a83 DB |
592 | __u8 meta_len; |
593 | __u8 nr_frags; | |
9f42f126 | 594 | __u8 tx_flags; |
7967168c HX |
595 | unsigned short gso_size; |
596 | /* Warning: this field is not always filled in (UFO)! */ | |
597 | unsigned short gso_segs; | |
1da177e4 | 598 | struct sk_buff *frag_list; |
48eb03dd SF |
599 | union { |
600 | struct skb_shared_hwtstamps hwtstamps; | |
601 | struct xsk_tx_metadata_compl xsk_meta; | |
602 | }; | |
7f564528 | 603 | unsigned int gso_type; |
09c2d251 | 604 | u32 tskey; |
ec7d2f2c ED |
605 | |
606 | /* | |
607 | * Warning : all fields before dataref are cleared in __alloc_skb() | |
608 | */ | |
609 | atomic_t dataref; | |
d16697cb | 610 | unsigned int xdp_frags_size; |
ec7d2f2c | 611 | |
69e3c75f JB |
612 | /* Intermediate layers must ensure that destructor_arg |
613 | * remains valid until skb destructor */ | |
614 | void * destructor_arg; | |
a6686f2f | 615 | |
fed66381 ED |
616 | /* must be last field, see pskb_expand_head() */ |
617 | skb_frag_t frags[MAX_SKB_FRAGS]; | |
1da177e4 LT |
618 | }; |
619 | ||
9ec7ea14 JK |
620 | /** |
621 | * DOC: dataref and headerless skbs | |
622 | * | |
623 | * Transport layers send out clones of payload skbs they hold for | |
624 | * retransmissions. To allow lower layers of the stack to prepend their headers | |
625 | * we split &skb_shared_info.dataref into two halves. | |
626 | * The lower 16 bits count the overall number of references. | |
627 | * The higher 16 bits indicate how many of the references are payload-only. | |
628 | * skb_header_cloned() checks if skb is allowed to add / write the headers. | |
629 | * | |
630 | * The creator of the skb (e.g. TCP) marks its skb as &sk_buff.nohdr | |
631 | * (via __skb_header_release()). Any clone created from marked skb will get | |
632 | * &sk_buff.hdr_len populated with the available headroom. | |
633 | * If there's the only clone in existence it's able to modify the headroom | |
634 | * at will. The sequence of calls inside the transport layer is:: | |
635 | * | |
636 | * <alloc skb> | |
637 | * skb_reserve() | |
638 | * __skb_header_release() | |
639 | * skb_clone() | |
640 | * // send the clone down the stack | |
641 | * | |
642 | * This is not a very generic construct and it depends on the transport layers | |
643 | * doing the right thing. In practice there's usually only one payload-only skb. | |
644 | * Having multiple payload-only skbs with different lengths of hdr_len is not | |
645 | * possible. The payload-only skbs should never leave their owner. | |
1da177e4 LT |
646 | */ |
647 | #define SKB_DATAREF_SHIFT 16 | |
648 | #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1) | |
649 | ||
d179cd12 DM |
650 | |
651 | enum { | |
c8753d55 VS |
652 | SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */ |
653 | SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */ | |
654 | SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */ | |
d179cd12 DM |
655 | }; |
656 | ||
7967168c HX |
657 | enum { |
658 | SKB_GSO_TCPV4 = 1 << 0, | |
576a30eb HX |
659 | |
660 | /* This indicates the skb is from an untrusted source. */ | |
d9d30adf | 661 | SKB_GSO_DODGY = 1 << 1, |
b0da8537 MC |
662 | |
663 | /* This indicates the tcp segment has CWR set. */ | |
d9d30adf | 664 | SKB_GSO_TCP_ECN = 1 << 2, |
f83ef8c0 | 665 | |
d9d30adf | 666 | SKB_GSO_TCP_FIXEDID = 1 << 3, |
01d5b2fc | 667 | |
d9d30adf | 668 | SKB_GSO_TCPV6 = 1 << 4, |
68c33163 | 669 | |
d9d30adf | 670 | SKB_GSO_FCOE = 1 << 5, |
73136267 | 671 | |
d9d30adf | 672 | SKB_GSO_GRE = 1 << 6, |
0d89d203 | 673 | |
d9d30adf | 674 | SKB_GSO_GRE_CSUM = 1 << 7, |
cb32f511 | 675 | |
d9d30adf | 676 | SKB_GSO_IPXIP4 = 1 << 8, |
61c1db7f | 677 | |
d9d30adf | 678 | SKB_GSO_IPXIP6 = 1 << 9, |
0f4f4ffa | 679 | |
d9d30adf | 680 | SKB_GSO_UDP_TUNNEL = 1 << 10, |
4749c09c | 681 | |
d9d30adf | 682 | SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11, |
cbc53e08 | 683 | |
d9d30adf | 684 | SKB_GSO_PARTIAL = 1 << 12, |
802ab55a | 685 | |
d9d30adf | 686 | SKB_GSO_TUNNEL_REMCSUM = 1 << 13, |
90017acc | 687 | |
d9d30adf | 688 | SKB_GSO_SCTP = 1 << 14, |
c7ef8f0c | 689 | |
d9d30adf | 690 | SKB_GSO_ESP = 1 << 15, |
0c19f846 WB |
691 | |
692 | SKB_GSO_UDP = 1 << 16, | |
ee80d1eb WB |
693 | |
694 | SKB_GSO_UDP_L4 = 1 << 17, | |
3b335832 SK |
695 | |
696 | SKB_GSO_FRAGLIST = 1 << 18, | |
7967168c HX |
697 | }; |
698 | ||
2e07fa9c ACM |
699 | #if BITS_PER_LONG > 32 |
700 | #define NET_SKBUFF_DATA_USES_OFFSET 1 | |
701 | #endif | |
702 | ||
703 | #ifdef NET_SKBUFF_DATA_USES_OFFSET | |
704 | typedef unsigned int sk_buff_data_t; | |
705 | #else | |
706 | typedef unsigned char *sk_buff_data_t; | |
707 | #endif | |
708 | ||
ddccc9ef JK |
709 | /** |
710 | * DOC: Basic sk_buff geometry | |
711 | * | |
712 | * struct sk_buff itself is a metadata structure and does not hold any packet | |
713 | * data. All the data is held in associated buffers. | |
714 | * | |
715 | * &sk_buff.head points to the main "head" buffer. The head buffer is divided | |
716 | * into two parts: | |
717 | * | |
718 | * - data buffer, containing headers and sometimes payload; | |
719 | * this is the part of the skb operated on by the common helpers | |
720 | * such as skb_put() or skb_pull(); | |
721 | * - shared info (struct skb_shared_info) which holds an array of pointers | |
722 | * to read-only data in the (page, offset, length) format. | |
723 | * | |
724 | * Optionally &skb_shared_info.frag_list may point to another skb. | |
725 | * | |
726 | * Basic diagram may look like this:: | |
727 | * | |
728 | * --------------- | |
729 | * | sk_buff | | |
730 | * --------------- | |
731 | * ,--------------------------- + head | |
732 | * / ,----------------- + data | |
733 | * / / ,----------- + tail | |
734 | * | | | , + end | |
735 | * | | | | | |
736 | * v v v v | |
737 | * ----------------------------------------------- | |
738 | * | headroom | data | tailroom | skb_shared_info | | |
739 | * ----------------------------------------------- | |
740 | * + [page frag] | |
741 | * + [page frag] | |
742 | * + [page frag] | |
743 | * + [page frag] --------- | |
744 | * + frag_list --> | sk_buff | | |
745 | * --------- | |
746 | * | |
747 | */ | |
748 | ||
161e6137 | 749 | /** |
1da177e4 LT |
750 | * struct sk_buff - socket buffer |
751 | * @next: Next buffer in list | |
752 | * @prev: Previous buffer in list | |
363ec392 | 753 | * @tstamp: Time we arrived/left |
d2f273f0 RD |
754 | * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point |
755 | * for retransmit timer | |
56b17425 | 756 | * @rbnode: RB tree node, alternative to next/prev for netem/tcp |
d2f273f0 | 757 | * @list: queue head |
df6160de | 758 | * @ll_node: anchor in an llist (eg socket defer_list) |
d84e0bd7 | 759 | * @sk: Socket we are owned by |
1da177e4 | 760 | * @dev: Device we arrived on/are leaving by |
d2f273f0 | 761 | * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL |
d84e0bd7 | 762 | * @cb: Control buffer. Free for use by every layer. Put private vars here |
7fee226a | 763 | * @_skb_refdst: destination entry (with norefcount bit) |
1da177e4 LT |
764 | * @len: Length of actual data |
765 | * @data_len: Data length | |
766 | * @mac_len: Length of link layer header | |
334a8132 | 767 | * @hdr_len: writable header length of cloned skb |
663ead3b HX |
768 | * @csum: Checksum (must include start/offset pair) |
769 | * @csum_start: Offset from skb->head where checksumming should start | |
770 | * @csum_offset: Offset from csum_start where checksum should be stored | |
d84e0bd7 | 771 | * @priority: Packet queueing priority |
60ff7467 | 772 | * @ignore_df: allow local fragmentation |
1da177e4 | 773 | * @cloned: Head may be cloned (check refcnt to be sure) |
d84e0bd7 | 774 | * @ip_summed: Driver fed us an IP checksum |
1da177e4 LT |
775 | * @nohdr: Payload reference only, must not modify header |
776 | * @pkt_type: Packet class | |
c83c2486 | 777 | * @fclone: skbuff clone status |
c83c2486 | 778 | * @ipvs_property: skbuff is owned by ipvs |
d2f273f0 RD |
779 | * @inner_protocol_type: whether the inner protocol is |
780 | * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO | |
781 | * @remcsum_offload: remote checksum offload is enabled | |
875e8939 IS |
782 | * @offload_fwd_mark: Packet was L2-forwarded in hardware |
783 | * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware | |
e7246e12 | 784 | * @tc_skip_classify: do not classify packet. set by IFB device |
8dc07fdb | 785 | * @tc_at_ingress: used within tc_classify to distinguish in/egress |
2c64605b PNA |
786 | * @redirected: packet was redirected by packet classifier |
787 | * @from_ingress: packet was redirected from the ingress path | |
42df6e1d | 788 | * @nf_skip_egress: packet shall skip nf egress - see netfilter_netdev.h |
31729363 RD |
789 | * @peeked: this packet has been seen already, so stats have been |
790 | * done for it, don't do them again | |
ba9dda3a | 791 | * @nf_trace: netfilter packet trace flag |
d84e0bd7 DB |
792 | * @protocol: Packet protocol from driver |
793 | * @destructor: Destruct function | |
e2080072 | 794 | * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue) |
6ed6e1c7 | 795 | * @_sk_redir: socket redirection information for skmsg |
a9e419dc | 796 | * @_nfct: Associated connection, if any (with nfctinfo bits) |
8964be4a | 797 | * @skb_iif: ifindex of device we arrived on |
1da177e4 | 798 | * @tc_index: Traffic control index |
61b905da | 799 | * @hash: the packet hash |
d84e0bd7 | 800 | * @queue_mapping: Queue mapping for multiqueue devices |
d2f273f0 RD |
801 | * @head_frag: skb was allocated from page fragments, |
802 | * not allocated by kmalloc() or vmalloc(). | |
8b700862 | 803 | * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves |
6a5bcd84 IA |
804 | * @pp_recycle: mark the packet for recycling instead of freeing (implies |
805 | * page_pool support on driver) | |
df5042f4 | 806 | * @active_extensions: active extensions (skb_ext_id types) |
553a5672 | 807 | * @ndisc_nodetype: router type (from link layer) |
d84e0bd7 | 808 | * @ooo_okay: allow the mapping of a socket to a queue to be changed |
61b905da | 809 | * @l4_hash: indicate hash is a canonical 4-tuple hash over transport |
4ca2462e | 810 | * ports. |
a3b18ddb | 811 | * @sw_hash: indicates hash was computed in software stack |
6e3e939f JB |
812 | * @wifi_acked_valid: wifi_acked was set |
813 | * @wifi_acked: whether frame was acked on wifi or not | |
3bdc0eba | 814 | * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS |
d2f273f0 RD |
815 | * @encapsulation: indicates the inner headers in the skbuff are valid |
816 | * @encap_hdr_csum: software checksum is needed | |
817 | * @csum_valid: checksum is already valid | |
dba00306 | 818 | * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL |
d2f273f0 RD |
819 | * @csum_complete_sw: checksum was completed by software |
820 | * @csum_level: indicates the number of consecutive checksums found in | |
821 | * the packet minus one that have been verified as | |
822 | * CHECKSUM_UNNECESSARY (max 3) | |
4ff06203 | 823 | * @dst_pending_confirm: need to confirm neighbour |
a48d189e | 824 | * @decrypted: Decrypted SKB |
5fc88f93 | 825 | * @slow_gro: state present at GRO time, slower prepare step required |
a1ac9c8a | 826 | * @mono_delivery_time: When set, skb->tstamp has the |
35c3e279 AC |
827 | * delivery_time in mono clock base (i.e. EDT). Otherwise, the |
828 | * skb->tstamp has the (rcv) timestamp at ingress and | |
829 | * delivery_time at egress. | |
161e6137 | 830 | * @napi_id: id of the NAPI struct this skb came from |
d2f273f0 | 831 | * @sender_cpu: (aka @napi_id) source CPU in XPS |
68822bdf | 832 | * @alloc_cpu: CPU which did the skb allocation. |
984bc16c | 833 | * @secmark: security marking |
d84e0bd7 | 834 | * @mark: Generic packet mark |
d2f273f0 RD |
835 | * @reserved_tailroom: (aka @mark) number of bytes of free space available |
836 | * at the tail of an sk_buff | |
354259fa | 837 | * @vlan_all: vlan fields (proto & tci) |
86a9bad3 | 838 | * @vlan_proto: vlan encapsulation protocol |
6aa895b0 | 839 | * @vlan_tci: vlan tag control information |
0d89d203 | 840 | * @inner_protocol: Protocol (encapsulation) |
d2f273f0 RD |
841 | * @inner_ipproto: (aka @inner_protocol) stores ipproto when |
842 | * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO; | |
6a674e9c JG |
843 | * @inner_transport_header: Inner transport layer header (encapsulation) |
844 | * @inner_network_header: Network layer header (encapsulation) | |
aefbd2b3 | 845 | * @inner_mac_header: Link layer header (encapsulation) |
d84e0bd7 DB |
846 | * @transport_header: Transport layer header |
847 | * @network_header: Network layer header | |
848 | * @mac_header: Link layer header | |
fa69ee5a | 849 | * @kcov_handle: KCOV remote handle for remote coverage collection |
d84e0bd7 DB |
850 | * @tail: Tail pointer |
851 | * @end: End pointer | |
852 | * @head: Head of buffer | |
853 | * @data: Data head pointer | |
854 | * @truesize: Buffer size | |
855 | * @users: User count - see {datagram,tcp}.c | |
df5042f4 | 856 | * @extensions: allocated extensions, valid if active_extensions is nonzero |
1da177e4 LT |
857 | */ |
858 | ||
859 | struct sk_buff { | |
363ec392 | 860 | union { |
56b17425 | 861 | struct { |
1a2fb220 | 862 | /* These two members must be first to match sk_buff_head. */ |
56b17425 ED |
863 | struct sk_buff *next; |
864 | struct sk_buff *prev; | |
865 | ||
866 | union { | |
bffa72cf ED |
867 | struct net_device *dev; |
868 | /* Some protocols might use this space to store information, | |
869 | * while device pointer would be NULL. | |
870 | * UDP receive path is one user. | |
871 | */ | |
872 | unsigned long dev_scratch; | |
56b17425 ED |
873 | }; |
874 | }; | |
fa0f5273 | 875 | struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */ |
d4546c25 | 876 | struct list_head list; |
f35f8219 | 877 | struct llist_node ll_node; |
363ec392 | 878 | }; |
fa0f5273 | 879 | |
18685451 | 880 | struct sock *sk; |
1da177e4 | 881 | |
c84d9490 | 882 | union { |
bffa72cf | 883 | ktime_t tstamp; |
d3edd06e | 884 | u64 skb_mstamp_ns; /* earliest departure time */ |
c84d9490 | 885 | }; |
1da177e4 LT |
886 | /* |
887 | * This is the control buffer. It is free to use for every | |
888 | * layer. Please put your private variables there. If you | |
889 | * want to keep them across layers you have to do a skb_clone() | |
890 | * first. This is owned by whoever has the skb queued ATM. | |
891 | */ | |
da3f5cf1 | 892 | char cb[48] __aligned(8); |
1da177e4 | 893 | |
e2080072 ED |
894 | union { |
895 | struct { | |
896 | unsigned long _skb_refdst; | |
897 | void (*destructor)(struct sk_buff *skb); | |
898 | }; | |
899 | struct list_head tcp_tsorted_anchor; | |
e3526bb9 CW |
900 | #ifdef CONFIG_NET_SOCK_MSG |
901 | unsigned long _sk_redir; | |
902 | #endif | |
e2080072 ED |
903 | }; |
904 | ||
b1937227 | 905 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) |
a9e419dc | 906 | unsigned long _nfct; |
da3f5cf1 | 907 | #endif |
1da177e4 | 908 | unsigned int len, |
334a8132 PM |
909 | data_len; |
910 | __u16 mac_len, | |
911 | hdr_len; | |
b1937227 ED |
912 | |
913 | /* Following fields are _not_ copied in __copy_skb_header() | |
914 | * Note that queue_mapping is here mostly to fill a hole. | |
915 | */ | |
b1937227 | 916 | __u16 queue_mapping; |
36bbef52 DB |
917 | |
918 | /* if you move cloned around you also must adapt those constants */ | |
919 | #ifdef __BIG_ENDIAN_BITFIELD | |
920 | #define CLONED_MASK (1 << 7) | |
921 | #else | |
922 | #define CLONED_MASK 1 | |
923 | #endif | |
fba84957 | 924 | #define CLONED_OFFSET offsetof(struct sk_buff, __cloned_offset) |
36bbef52 | 925 | |
d2f273f0 | 926 | /* private: */ |
36bbef52 | 927 | __u8 __cloned_offset[0]; |
d2f273f0 | 928 | /* public: */ |
b1937227 | 929 | __u8 cloned:1, |
6869c4d8 | 930 | nohdr:1, |
b84f4cc9 | 931 | fclone:2, |
a59322be | 932 | peeked:1, |
b1937227 | 933 | head_frag:1, |
6a5bcd84 IA |
934 | pfmemalloc:1, |
935 | pp_recycle:1; /* page_pool recycle indicator */ | |
df5042f4 FW |
936 | #ifdef CONFIG_SKB_EXTENSIONS |
937 | __u8 active_extensions; | |
938 | #endif | |
6a5bcd84 | 939 | |
03f61041 | 940 | /* Fields enclosed in headers group are copied |
b1937227 ED |
941 | * using a single memcpy() in __copy_skb_header() |
942 | */ | |
03f61041 | 943 | struct_group(headers, |
4031ae6e | 944 | |
d2f273f0 | 945 | /* private: */ |
233577a2 | 946 | __u8 __pkt_type_offset[0]; |
d2f273f0 | 947 | /* public: */ |
fba84957 | 948 | __u8 pkt_type:3; /* see PKT_TYPE_MAX */ |
b1937227 | 949 | __u8 ignore_df:1; |
4c60d04c | 950 | __u8 dst_pending_confirm:1; |
b1937227 | 951 | __u8 ip_summed:2; |
3853b584 | 952 | __u8 ooo_okay:1; |
8b700862 | 953 | |
d2f273f0 | 954 | /* private: */ |
04aae213 | 955 | __u8 __mono_tc_offset[0]; |
d2f273f0 | 956 | /* public: */ |
3b5d4ddf | 957 | __u8 mono_delivery_time:1; /* See SKB_MONO_DELIVERY_TIME_MASK */ |
e420bed0 | 958 | #ifdef CONFIG_NET_XGRESS |
3b5d4ddf | 959 | __u8 tc_at_ingress:1; /* See TC_AT_INGRESS_MASK */ |
c0ba8611 | 960 | __u8 tc_skip_classify:1; |
7449197d | 961 | #endif |
c0ba8611 JK |
962 | __u8 remcsum_offload:1; |
963 | __u8 csum_complete_sw:1; | |
964 | __u8 csum_level:2; | |
4c60d04c | 965 | __u8 inner_protocol_type:1; |
b94e032b | 966 | |
61b905da | 967 | __u8 l4_hash:1; |
a3b18ddb | 968 | __u8 sw_hash:1; |
eb6fba75 | 969 | #ifdef CONFIG_WIRELESS |
6e3e939f JB |
970 | __u8 wifi_acked_valid:1; |
971 | __u8 wifi_acked:1; | |
eb6fba75 | 972 | #endif |
3bdc0eba | 973 | __u8 no_fcs:1; |
77cffe23 | 974 | /* Indicates the inner headers are valid in the skbuff. */ |
6a674e9c | 975 | __u8 encapsulation:1; |
7e2b10c1 | 976 | __u8 encap_hdr_csum:1; |
5d0c2b95 | 977 | __u8 csum_valid:1; |
b1937227 ED |
978 | #ifdef CONFIG_IPV6_NDISC_NODETYPE |
979 | __u8 ndisc_nodetype:2; | |
980 | #endif | |
8b700862 | 981 | |
48d80c39 | 982 | #if IS_ENABLED(CONFIG_IP_VS) |
0c4b2d37 | 983 | __u8 ipvs_property:1; |
48d80c39 JK |
984 | #endif |
985 | #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || IS_ENABLED(CONFIG_NF_TABLES) | |
4c60d04c | 986 | __u8 nf_trace:1; |
48d80c39 | 987 | #endif |
6bc506b4 IS |
988 | #ifdef CONFIG_NET_SWITCHDEV |
989 | __u8 offload_fwd_mark:1; | |
875e8939 | 990 | __u8 offload_l3_fwd_mark:1; |
2c64605b | 991 | #endif |
2c64605b | 992 | __u8 redirected:1; |
11941f8a | 993 | #ifdef CONFIG_NET_REDIRECT |
2c64605b | 994 | __u8 from_ingress:1; |
e7246e12 | 995 | #endif |
42df6e1d LW |
996 | #ifdef CONFIG_NETFILTER_SKIP_EGRESS |
997 | __u8 nf_skip_egress:1; | |
998 | #endif | |
9f06f87f | 999 | #ifdef CONFIG_SKB_DECRYPTED |
a48d189e SB |
1000 | __u8 decrypted:1; |
1001 | #endif | |
5fc88f93 | 1002 | __u8 slow_gro:1; |
c24831a1 | 1003 | #if IS_ENABLED(CONFIG_IP_SCTP) |
7449197d | 1004 | __u8 csum_not_inet:1; |
c24831a1 | 1005 | #endif |
b1937227 | 1006 | |
e420bed0 | 1007 | #if defined(CONFIG_NET_SCHED) || defined(CONFIG_NET_XGRESS) |
b1937227 | 1008 | __u16 tc_index; /* traffic control index */ |
b1937227 | 1009 | #endif |
fe55f6d5 | 1010 | |
4398f3f6 JK |
1011 | u16 alloc_cpu; |
1012 | ||
b1937227 ED |
1013 | union { |
1014 | __wsum csum; | |
1015 | struct { | |
1016 | __u16 csum_start; | |
1017 | __u16 csum_offset; | |
1018 | }; | |
1019 | }; | |
1020 | __u32 priority; | |
1021 | int skb_iif; | |
1022 | __u32 hash; | |
354259fa ED |
1023 | union { |
1024 | u32 vlan_all; | |
1025 | struct { | |
1026 | __be16 vlan_proto; | |
1027 | __u16 vlan_tci; | |
1028 | }; | |
1029 | }; | |
2bd82484 ED |
1030 | #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS) |
1031 | union { | |
1032 | unsigned int napi_id; | |
1033 | unsigned int sender_cpu; | |
1034 | }; | |
97fc2f08 | 1035 | #endif |
984bc16c | 1036 | #ifdef CONFIG_NETWORK_SECMARK |
6bc506b4 | 1037 | __u32 secmark; |
0c4f691f | 1038 | #endif |
0c4f691f | 1039 | |
3b885787 NH |
1040 | union { |
1041 | __u32 mark; | |
16fad69c | 1042 | __u32 reserved_tailroom; |
3b885787 | 1043 | }; |
1da177e4 | 1044 | |
8bce6d7d TH |
1045 | union { |
1046 | __be16 inner_protocol; | |
1047 | __u8 inner_ipproto; | |
1048 | }; | |
1049 | ||
1a37e412 SH |
1050 | __u16 inner_transport_header; |
1051 | __u16 inner_network_header; | |
1052 | __u16 inner_mac_header; | |
b1937227 ED |
1053 | |
1054 | __be16 protocol; | |
1a37e412 SH |
1055 | __u16 transport_header; |
1056 | __u16 network_header; | |
1057 | __u16 mac_header; | |
b1937227 | 1058 | |
fa69ee5a ME |
1059 | #ifdef CONFIG_KCOV |
1060 | u64 kcov_handle; | |
1061 | #endif | |
1062 | ||
03f61041 | 1063 | ); /* end headers group */ |
b1937227 | 1064 | |
1da177e4 | 1065 | /* These elements must be at the end, see alloc_skb() for details. */ |
27a884dc | 1066 | sk_buff_data_t tail; |
4305b541 | 1067 | sk_buff_data_t end; |
1da177e4 | 1068 | unsigned char *head, |
4305b541 | 1069 | *data; |
27a884dc | 1070 | unsigned int truesize; |
63354797 | 1071 | refcount_t users; |
df5042f4 FW |
1072 | |
1073 | #ifdef CONFIG_SKB_EXTENSIONS | |
bf873a80 | 1074 | /* only usable after checking ->active_extensions != 0 */ |
df5042f4 FW |
1075 | struct skb_ext *extensions; |
1076 | #endif | |
1da177e4 LT |
1077 | }; |
1078 | ||
fba84957 KC |
1079 | /* if you move pkt_type around you also must adapt those constants */ |
1080 | #ifdef __BIG_ENDIAN_BITFIELD | |
1081 | #define PKT_TYPE_MAX (7 << 5) | |
1082 | #else | |
1083 | #define PKT_TYPE_MAX 7 | |
1084 | #endif | |
1085 | #define PKT_TYPE_OFFSET offsetof(struct sk_buff, __pkt_type_offset) | |
1086 | ||
354259fa | 1087 | /* if you move tc_at_ingress or mono_delivery_time |
3b5d4ddf MKL |
1088 | * around, you also must adapt these constants. |
1089 | */ | |
fba84957 | 1090 | #ifdef __BIG_ENDIAN_BITFIELD |
c0ba8611 JK |
1091 | #define SKB_MONO_DELIVERY_TIME_MASK (1 << 7) |
1092 | #define TC_AT_INGRESS_MASK (1 << 6) | |
fba84957 | 1093 | #else |
c0ba8611 JK |
1094 | #define SKB_MONO_DELIVERY_TIME_MASK (1 << 0) |
1095 | #define TC_AT_INGRESS_MASK (1 << 1) | |
fba84957 | 1096 | #endif |
04aae213 | 1097 | #define SKB_BF_MONO_TC_OFFSET offsetof(struct sk_buff, __mono_tc_offset) |
fba84957 | 1098 | |
1da177e4 LT |
1099 | #ifdef __KERNEL__ |
1100 | /* | |
1101 | * Handling routines are only of interest to the kernel | |
1102 | */ | |
1da177e4 | 1103 | |
c93bdd0e MG |
1104 | #define SKB_ALLOC_FCLONE 0x01 |
1105 | #define SKB_ALLOC_RX 0x02 | |
fd11a83d | 1106 | #define SKB_ALLOC_NAPI 0x04 |
c93bdd0e | 1107 | |
161e6137 PT |
1108 | /** |
1109 | * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves | |
1110 | * @skb: buffer | |
1111 | */ | |
c93bdd0e MG |
1112 | static inline bool skb_pfmemalloc(const struct sk_buff *skb) |
1113 | { | |
1114 | return unlikely(skb->pfmemalloc); | |
1115 | } | |
1116 | ||
7fee226a ED |
1117 | /* |
1118 | * skb might have a dst pointer attached, refcounted or not. | |
1119 | * _skb_refdst low order bit is set if refcount was _not_ taken | |
1120 | */ | |
1121 | #define SKB_DST_NOREF 1UL | |
1122 | #define SKB_DST_PTRMASK ~(SKB_DST_NOREF) | |
1123 | ||
1124 | /** | |
1125 | * skb_dst - returns skb dst_entry | |
1126 | * @skb: buffer | |
1127 | * | |
1128 | * Returns skb dst_entry, regardless of reference taken or not. | |
1129 | */ | |
adf30907 ED |
1130 | static inline struct dst_entry *skb_dst(const struct sk_buff *skb) |
1131 | { | |
161e6137 | 1132 | /* If refdst was not refcounted, check we still are in a |
7fee226a ED |
1133 | * rcu_read_lock section |
1134 | */ | |
1135 | WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) && | |
1136 | !rcu_read_lock_held() && | |
1137 | !rcu_read_lock_bh_held()); | |
1138 | return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK); | |
adf30907 ED |
1139 | } |
1140 | ||
7fee226a ED |
1141 | /** |
1142 | * skb_dst_set - sets skb dst | |
1143 | * @skb: buffer | |
1144 | * @dst: dst entry | |
1145 | * | |
1146 | * Sets skb dst, assuming a reference was taken on dst and should | |
1147 | * be released by skb_dst_drop() | |
1148 | */ | |
adf30907 ED |
1149 | static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst) |
1150 | { | |
8a886b14 | 1151 | skb->slow_gro |= !!dst; |
7fee226a ED |
1152 | skb->_skb_refdst = (unsigned long)dst; |
1153 | } | |
1154 | ||
932bc4d7 JA |
1155 | /** |
1156 | * skb_dst_set_noref - sets skb dst, hopefully, without taking reference | |
1157 | * @skb: buffer | |
1158 | * @dst: dst entry | |
1159 | * | |
1160 | * Sets skb dst, assuming a reference was not taken on dst. | |
1161 | * If dst entry is cached, we do not take reference and dst_release | |
1162 | * will be avoided by refdst_drop. If dst entry is not cached, we take | |
1163 | * reference, so that last dst_release can destroy the dst immediately. | |
1164 | */ | |
1165 | static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst) | |
1166 | { | |
dbfc4fb7 | 1167 | WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held()); |
a432934a | 1168 | skb->slow_gro |= !!dst; |
dbfc4fb7 | 1169 | skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF; |
932bc4d7 | 1170 | } |
7fee226a ED |
1171 | |
1172 | /** | |
25985edc | 1173 | * skb_dst_is_noref - Test if skb dst isn't refcounted |
7fee226a ED |
1174 | * @skb: buffer |
1175 | */ | |
1176 | static inline bool skb_dst_is_noref(const struct sk_buff *skb) | |
1177 | { | |
1178 | return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb); | |
adf30907 ED |
1179 | } |
1180 | ||
8b10cab6 JHS |
1181 | /* For mangling skb->pkt_type from user space side from applications |
1182 | * such as nft, tc, etc, we only allow a conservative subset of | |
1183 | * possible pkt_types to be set. | |
1184 | */ | |
1185 | static inline bool skb_pkt_type_ok(u32 ptype) | |
1186 | { | |
1187 | return ptype <= PACKET_OTHERHOST; | |
1188 | } | |
1189 | ||
161e6137 PT |
1190 | /** |
1191 | * skb_napi_id - Returns the skb's NAPI id | |
1192 | * @skb: buffer | |
1193 | */ | |
90b602f8 ML |
1194 | static inline unsigned int skb_napi_id(const struct sk_buff *skb) |
1195 | { | |
1196 | #ifdef CONFIG_NET_RX_BUSY_POLL | |
1197 | return skb->napi_id; | |
1198 | #else | |
1199 | return 0; | |
eb6fba75 JK |
1200 | #endif |
1201 | } | |
1202 | ||
1203 | static inline bool skb_wifi_acked_valid(const struct sk_buff *skb) | |
1204 | { | |
1205 | #ifdef CONFIG_WIRELESS | |
1206 | return skb->wifi_acked_valid; | |
1207 | #else | |
1208 | return 0; | |
90b602f8 ML |
1209 | #endif |
1210 | } | |
1211 | ||
161e6137 PT |
1212 | /** |
1213 | * skb_unref - decrement the skb's reference count | |
1214 | * @skb: buffer | |
1215 | * | |
1216 | * Returns true if we can free the skb. | |
1217 | */ | |
3889a803 PA |
1218 | static inline bool skb_unref(struct sk_buff *skb) |
1219 | { | |
1220 | if (unlikely(!skb)) | |
1221 | return false; | |
63354797 | 1222 | if (likely(refcount_read(&skb->users) == 1)) |
3889a803 | 1223 | smp_rmb(); |
63354797 | 1224 | else if (likely(!refcount_dec_and_test(&skb->users))) |
3889a803 PA |
1225 | return false; |
1226 | ||
1227 | return true; | |
1228 | } | |
1229 | ||
1cface55 ED |
1230 | static inline bool skb_data_unref(const struct sk_buff *skb, |
1231 | struct skb_shared_info *shinfo) | |
1232 | { | |
1233 | int bias; | |
1234 | ||
1235 | if (!skb->cloned) | |
1236 | return true; | |
1237 | ||
1238 | bias = skb->nohdr ? (1 << SKB_DATAREF_SHIFT) + 1 : 1; | |
1239 | ||
1240 | if (atomic_read(&shinfo->dataref) == bias) | |
1241 | smp_rmb(); | |
1242 | else if (atomic_sub_return(bias, &shinfo->dataref)) | |
1243 | return false; | |
1244 | ||
1245 | return true; | |
1246 | } | |
1247 | ||
c205cc75 MD |
1248 | void __fix_address |
1249 | kfree_skb_reason(struct sk_buff *skb, enum skb_drop_reason reason); | |
c504e5c2 MD |
1250 | |
1251 | /** | |
1252 | * kfree_skb - free an sk_buff with 'NOT_SPECIFIED' reason | |
1253 | * @skb: buffer to free | |
1254 | */ | |
1255 | static inline void kfree_skb(struct sk_buff *skb) | |
1256 | { | |
1257 | kfree_skb_reason(skb, SKB_DROP_REASON_NOT_SPECIFIED); | |
1258 | } | |
1259 | ||
0a463c78 | 1260 | void skb_release_head_state(struct sk_buff *skb); |
215b0f19 MD |
1261 | void kfree_skb_list_reason(struct sk_buff *segs, |
1262 | enum skb_drop_reason reason); | |
6413139d | 1263 | void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt); |
7965bd4d | 1264 | void skb_tx_error(struct sk_buff *skb); |
be769db2 | 1265 | |
215b0f19 MD |
1266 | static inline void kfree_skb_list(struct sk_buff *segs) |
1267 | { | |
1268 | kfree_skb_list_reason(segs, SKB_DROP_REASON_NOT_SPECIFIED); | |
1269 | } | |
1270 | ||
be769db2 | 1271 | #ifdef CONFIG_TRACEPOINTS |
7965bd4d | 1272 | void consume_skb(struct sk_buff *skb); |
be769db2 HX |
1273 | #else |
1274 | static inline void consume_skb(struct sk_buff *skb) | |
1275 | { | |
1276 | return kfree_skb(skb); | |
1277 | } | |
1278 | #endif | |
1279 | ||
ca2c1418 | 1280 | void __consume_stateless_skb(struct sk_buff *skb); |
7965bd4d | 1281 | void __kfree_skb(struct sk_buff *skb); |
bad43ca8 | 1282 | |
7965bd4d JP |
1283 | void kfree_skb_partial(struct sk_buff *skb, bool head_stolen); |
1284 | bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from, | |
1285 | bool *fragstolen, int *delta_truesize); | |
bad43ca8 | 1286 | |
7965bd4d JP |
1287 | struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags, |
1288 | int node); | |
2ea2f62c | 1289 | struct sk_buff *__build_skb(void *data, unsigned int frag_size); |
7965bd4d | 1290 | struct sk_buff *build_skb(void *data, unsigned int frag_size); |
ba0509b6 JDB |
1291 | struct sk_buff *build_skb_around(struct sk_buff *skb, |
1292 | void *data, unsigned int frag_size); | |
68822bdf | 1293 | void skb_attempt_defer_free(struct sk_buff *skb); |
161e6137 | 1294 | |
f450d539 | 1295 | struct sk_buff *napi_build_skb(void *data, unsigned int frag_size); |
ce098da1 | 1296 | struct sk_buff *slab_build_skb(void *data); |
f450d539 | 1297 | |
161e6137 PT |
1298 | /** |
1299 | * alloc_skb - allocate a network buffer | |
1300 | * @size: size to allocate | |
1301 | * @priority: allocation mask | |
1302 | * | |
1303 | * This function is a convenient wrapper around __alloc_skb(). | |
1304 | */ | |
d179cd12 | 1305 | static inline struct sk_buff *alloc_skb(unsigned int size, |
dd0fc66f | 1306 | gfp_t priority) |
d179cd12 | 1307 | { |
564824b0 | 1308 | return __alloc_skb(size, priority, 0, NUMA_NO_NODE); |
d179cd12 DM |
1309 | } |
1310 | ||
2e4e4410 ED |
1311 | struct sk_buff *alloc_skb_with_frags(unsigned long header_len, |
1312 | unsigned long data_len, | |
1313 | int max_page_order, | |
1314 | int *errcode, | |
1315 | gfp_t gfp_mask); | |
da29e4b4 | 1316 | struct sk_buff *alloc_skb_for_msg(struct sk_buff *first); |
2e4e4410 | 1317 | |
d0bf4a9e ED |
1318 | /* Layout of fast clones : [skb1][skb2][fclone_ref] */ |
1319 | struct sk_buff_fclones { | |
1320 | struct sk_buff skb1; | |
1321 | ||
1322 | struct sk_buff skb2; | |
1323 | ||
2638595a | 1324 | refcount_t fclone_ref; |
d0bf4a9e ED |
1325 | }; |
1326 | ||
1327 | /** | |
1328 | * skb_fclone_busy - check if fclone is busy | |
293de7de | 1329 | * @sk: socket |
d0bf4a9e ED |
1330 | * @skb: buffer |
1331 | * | |
bda13fed | 1332 | * Returns true if skb is a fast clone, and its clone is not freed. |
39bb5e62 | 1333 | * Some drivers call skb_orphan() in their ndo_start_xmit(), |
8527ca77 | 1334 | * so we also check that didn't happen. |
d0bf4a9e | 1335 | */ |
39bb5e62 ED |
1336 | static inline bool skb_fclone_busy(const struct sock *sk, |
1337 | const struct sk_buff *skb) | |
d0bf4a9e ED |
1338 | { |
1339 | const struct sk_buff_fclones *fclones; | |
1340 | ||
1341 | fclones = container_of(skb, struct sk_buff_fclones, skb1); | |
1342 | ||
1343 | return skb->fclone == SKB_FCLONE_ORIG && | |
2638595a | 1344 | refcount_read(&fclones->fclone_ref) > 1 && |
f4dae54e | 1345 | READ_ONCE(fclones->skb2.sk) == sk; |
d0bf4a9e ED |
1346 | } |
1347 | ||
161e6137 PT |
1348 | /** |
1349 | * alloc_skb_fclone - allocate a network buffer from fclone cache | |
1350 | * @size: size to allocate | |
1351 | * @priority: allocation mask | |
1352 | * | |
1353 | * This function is a convenient wrapper around __alloc_skb(). | |
1354 | */ | |
d179cd12 | 1355 | static inline struct sk_buff *alloc_skb_fclone(unsigned int size, |
dd0fc66f | 1356 | gfp_t priority) |
d179cd12 | 1357 | { |
c93bdd0e | 1358 | return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE); |
d179cd12 DM |
1359 | } |
1360 | ||
7965bd4d | 1361 | struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src); |
b0768a86 | 1362 | void skb_headers_offset_update(struct sk_buff *skb, int off); |
7965bd4d JP |
1363 | int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask); |
1364 | struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority); | |
08303c18 | 1365 | void skb_copy_header(struct sk_buff *new, const struct sk_buff *old); |
7965bd4d | 1366 | struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority); |
bad93e9d OP |
1367 | struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom, |
1368 | gfp_t gfp_mask, bool fclone); | |
1369 | static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom, | |
1370 | gfp_t gfp_mask) | |
1371 | { | |
1372 | return __pskb_copy_fclone(skb, headroom, gfp_mask, false); | |
1373 | } | |
7965bd4d JP |
1374 | |
1375 | int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask); | |
1376 | struct sk_buff *skb_realloc_headroom(struct sk_buff *skb, | |
1377 | unsigned int headroom); | |
f1260ff1 | 1378 | struct sk_buff *skb_expand_head(struct sk_buff *skb, unsigned int headroom); |
7965bd4d JP |
1379 | struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom, |
1380 | int newtailroom, gfp_t priority); | |
48a1df65 JD |
1381 | int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg, |
1382 | int offset, int len); | |
1383 | int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, | |
1384 | int offset, int len); | |
7965bd4d | 1385 | int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer); |
cd0a137a FF |
1386 | int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error); |
1387 | ||
1388 | /** | |
1389 | * skb_pad - zero pad the tail of an skb | |
1390 | * @skb: buffer to pad | |
1391 | * @pad: space to pad | |
1392 | * | |
1393 | * Ensure that a buffer is followed by a padding area that is zero | |
1394 | * filled. Used by network drivers which may DMA or transfer data | |
1395 | * beyond the buffer end onto the wire. | |
1396 | * | |
1397 | * May return error in out of memory cases. The skb is freed on error. | |
1398 | */ | |
1399 | static inline int skb_pad(struct sk_buff *skb, int pad) | |
1400 | { | |
1401 | return __skb_pad(skb, pad, true); | |
1402 | } | |
ead2ceb0 | 1403 | #define dev_kfree_skb(a) consume_skb(a) |
1da177e4 | 1404 | |
be12a1fe | 1405 | int skb_append_pagefrags(struct sk_buff *skb, struct page *page, |
96449f90 | 1406 | int offset, size_t size, size_t max_frags); |
be12a1fe | 1407 | |
d94d9fee | 1408 | struct skb_seq_state { |
677e90ed TG |
1409 | __u32 lower_offset; |
1410 | __u32 upper_offset; | |
1411 | __u32 frag_idx; | |
1412 | __u32 stepped_offset; | |
1413 | struct sk_buff *root_skb; | |
1414 | struct sk_buff *cur_skb; | |
1415 | __u8 *frag_data; | |
97550f6f | 1416 | __u32 frag_off; |
677e90ed TG |
1417 | }; |
1418 | ||
7965bd4d JP |
1419 | void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from, |
1420 | unsigned int to, struct skb_seq_state *st); | |
1421 | unsigned int skb_seq_read(unsigned int consumed, const u8 **data, | |
1422 | struct skb_seq_state *st); | |
1423 | void skb_abort_seq_read(struct skb_seq_state *st); | |
677e90ed | 1424 | |
7965bd4d | 1425 | unsigned int skb_find_text(struct sk_buff *skb, unsigned int from, |
059a2440 | 1426 | unsigned int to, struct ts_config *config); |
3fc7e8a6 | 1427 | |
09323cc4 TH |
1428 | /* |
1429 | * Packet hash types specify the type of hash in skb_set_hash. | |
1430 | * | |
1431 | * Hash types refer to the protocol layer addresses which are used to | |
1432 | * construct a packet's hash. The hashes are used to differentiate or identify | |
1433 | * flows of the protocol layer for the hash type. Hash types are either | |
1434 | * layer-2 (L2), layer-3 (L3), or layer-4 (L4). | |
1435 | * | |
1436 | * Properties of hashes: | |
1437 | * | |
1438 | * 1) Two packets in different flows have different hash values | |
1439 | * 2) Two packets in the same flow should have the same hash value | |
1440 | * | |
1441 | * A hash at a higher layer is considered to be more specific. A driver should | |
1442 | * set the most specific hash possible. | |
1443 | * | |
1444 | * A driver cannot indicate a more specific hash than the layer at which a hash | |
1445 | * was computed. For instance an L3 hash cannot be set as an L4 hash. | |
1446 | * | |
1447 | * A driver may indicate a hash level which is less specific than the | |
1448 | * actual layer the hash was computed on. For instance, a hash computed | |
1449 | * at L4 may be considered an L3 hash. This should only be done if the | |
1450 | * driver can't unambiguously determine that the HW computed the hash at | |
1451 | * the higher layer. Note that the "should" in the second property above | |
1452 | * permits this. | |
1453 | */ | |
1454 | enum pkt_hash_types { | |
1455 | PKT_HASH_TYPE_NONE, /* Undefined type */ | |
1456 | PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */ | |
1457 | PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */ | |
1458 | PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */ | |
1459 | }; | |
1460 | ||
bcc83839 | 1461 | static inline void skb_clear_hash(struct sk_buff *skb) |
09323cc4 | 1462 | { |
bcc83839 | 1463 | skb->hash = 0; |
a3b18ddb | 1464 | skb->sw_hash = 0; |
bcc83839 TH |
1465 | skb->l4_hash = 0; |
1466 | } | |
1467 | ||
1468 | static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb) | |
1469 | { | |
1470 | if (!skb->l4_hash) | |
1471 | skb_clear_hash(skb); | |
1472 | } | |
1473 | ||
1474 | static inline void | |
1475 | __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4) | |
1476 | { | |
1477 | skb->l4_hash = is_l4; | |
1478 | skb->sw_hash = is_sw; | |
61b905da | 1479 | skb->hash = hash; |
09323cc4 TH |
1480 | } |
1481 | ||
bcc83839 TH |
1482 | static inline void |
1483 | skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type) | |
1484 | { | |
1485 | /* Used by drivers to set hash from HW */ | |
1486 | __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4); | |
1487 | } | |
1488 | ||
1489 | static inline void | |
1490 | __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4) | |
1491 | { | |
1492 | __skb_set_hash(skb, hash, true, is_l4); | |
1493 | } | |
1494 | ||
e5276937 | 1495 | void __skb_get_hash(struct sk_buff *skb); |
b917783c | 1496 | u32 __skb_get_hash_symmetric(const struct sk_buff *skb); |
e5276937 | 1497 | u32 skb_get_poff(const struct sk_buff *skb); |
f96533cd | 1498 | u32 __skb_get_poff(const struct sk_buff *skb, const void *data, |
72a338bc | 1499 | const struct flow_keys_basic *keys, int hlen); |
e5276937 | 1500 | __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto, |
f96533cd | 1501 | const void *data, int hlen_proto); |
e5276937 TH |
1502 | |
1503 | static inline __be32 skb_flow_get_ports(const struct sk_buff *skb, | |
1504 | int thoff, u8 ip_proto) | |
1505 | { | |
1506 | return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0); | |
1507 | } | |
1508 | ||
1509 | void skb_flow_dissector_init(struct flow_dissector *flow_dissector, | |
1510 | const struct flow_dissector_key *key, | |
1511 | unsigned int key_count); | |
1512 | ||
089b19a9 | 1513 | struct bpf_flow_dissector; |
0ba98502 SL |
1514 | u32 bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx, |
1515 | __be16 proto, int nhoff, int hlen, unsigned int flags); | |
089b19a9 | 1516 | |
3cbf4ffb SF |
1517 | bool __skb_flow_dissect(const struct net *net, |
1518 | const struct sk_buff *skb, | |
e5276937 | 1519 | struct flow_dissector *flow_dissector, |
f96533cd AL |
1520 | void *target_container, const void *data, |
1521 | __be16 proto, int nhoff, int hlen, unsigned int flags); | |
e5276937 TH |
1522 | |
1523 | static inline bool skb_flow_dissect(const struct sk_buff *skb, | |
1524 | struct flow_dissector *flow_dissector, | |
cd79a238 | 1525 | void *target_container, unsigned int flags) |
e5276937 | 1526 | { |
3cbf4ffb SF |
1527 | return __skb_flow_dissect(NULL, skb, flow_dissector, |
1528 | target_container, NULL, 0, 0, 0, flags); | |
e5276937 TH |
1529 | } |
1530 | ||
1531 | static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb, | |
cd79a238 TH |
1532 | struct flow_keys *flow, |
1533 | unsigned int flags) | |
e5276937 TH |
1534 | { |
1535 | memset(flow, 0, sizeof(*flow)); | |
3cbf4ffb SF |
1536 | return __skb_flow_dissect(NULL, skb, &flow_keys_dissector, |
1537 | flow, NULL, 0, 0, 0, flags); | |
e5276937 TH |
1538 | } |
1539 | ||
72a338bc | 1540 | static inline bool |
3cbf4ffb SF |
1541 | skb_flow_dissect_flow_keys_basic(const struct net *net, |
1542 | const struct sk_buff *skb, | |
f96533cd AL |
1543 | struct flow_keys_basic *flow, |
1544 | const void *data, __be16 proto, | |
1545 | int nhoff, int hlen, unsigned int flags) | |
e5276937 TH |
1546 | { |
1547 | memset(flow, 0, sizeof(*flow)); | |
3cbf4ffb | 1548 | return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow, |
cd79a238 | 1549 | data, proto, nhoff, hlen, flags); |
e5276937 TH |
1550 | } |
1551 | ||
82828b88 JP |
1552 | void skb_flow_dissect_meta(const struct sk_buff *skb, |
1553 | struct flow_dissector *flow_dissector, | |
1554 | void *target_container); | |
1555 | ||
75a56758 | 1556 | /* Gets a skb connection tracking info, ctinfo map should be a |
2ff17117 | 1557 | * map of mapsize to translate enum ip_conntrack_info states |
75a56758 PB |
1558 | * to user states. |
1559 | */ | |
1560 | void | |
1561 | skb_flow_dissect_ct(const struct sk_buff *skb, | |
1562 | struct flow_dissector *flow_dissector, | |
1563 | void *target_container, | |
7baf2429 | 1564 | u16 *ctinfo_map, size_t mapsize, |
38495958 | 1565 | bool post_ct, u16 zone); |
62b32379 SH |
1566 | void |
1567 | skb_flow_dissect_tunnel_info(const struct sk_buff *skb, | |
1568 | struct flow_dissector *flow_dissector, | |
1569 | void *target_container); | |
1570 | ||
0cb09aff AL |
1571 | void skb_flow_dissect_hash(const struct sk_buff *skb, |
1572 | struct flow_dissector *flow_dissector, | |
1573 | void *target_container); | |
1574 | ||
3958afa1 | 1575 | static inline __u32 skb_get_hash(struct sk_buff *skb) |
bfb564e7 | 1576 | { |
a3b18ddb | 1577 | if (!skb->l4_hash && !skb->sw_hash) |
3958afa1 | 1578 | __skb_get_hash(skb); |
bfb564e7 | 1579 | |
61b905da | 1580 | return skb->hash; |
bfb564e7 KK |
1581 | } |
1582 | ||
20a17bf6 | 1583 | static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6) |
f70ea018 | 1584 | { |
c6cc1ca7 TH |
1585 | if (!skb->l4_hash && !skb->sw_hash) { |
1586 | struct flow_keys keys; | |
de4c1f8b | 1587 | __u32 hash = __get_hash_from_flowi6(fl6, &keys); |
c6cc1ca7 | 1588 | |
de4c1f8b | 1589 | __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys)); |
c6cc1ca7 | 1590 | } |
f70ea018 TH |
1591 | |
1592 | return skb->hash; | |
1593 | } | |
1594 | ||
55667441 ED |
1595 | __u32 skb_get_hash_perturb(const struct sk_buff *skb, |
1596 | const siphash_key_t *perturb); | |
50fb7992 | 1597 | |
57bdf7f4 TH |
1598 | static inline __u32 skb_get_hash_raw(const struct sk_buff *skb) |
1599 | { | |
61b905da | 1600 | return skb->hash; |
57bdf7f4 TH |
1601 | } |
1602 | ||
3df7a74e TH |
1603 | static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from) |
1604 | { | |
61b905da | 1605 | to->hash = from->hash; |
a3b18ddb | 1606 | to->sw_hash = from->sw_hash; |
61b905da | 1607 | to->l4_hash = from->l4_hash; |
3df7a74e TH |
1608 | }; |
1609 | ||
14c4be92 JK |
1610 | static inline int skb_cmp_decrypted(const struct sk_buff *skb1, |
1611 | const struct sk_buff *skb2) | |
1612 | { | |
9f06f87f | 1613 | #ifdef CONFIG_SKB_DECRYPTED |
14c4be92 JK |
1614 | return skb2->decrypted - skb1->decrypted; |
1615 | #else | |
1616 | return 0; | |
1617 | #endif | |
1618 | } | |
1619 | ||
9f06f87f JK |
1620 | static inline bool skb_is_decrypted(const struct sk_buff *skb) |
1621 | { | |
1622 | #ifdef CONFIG_SKB_DECRYPTED | |
1623 | return skb->decrypted; | |
1624 | #else | |
1625 | return false; | |
1626 | #endif | |
1627 | } | |
1628 | ||
41477662 JK |
1629 | static inline void skb_copy_decrypted(struct sk_buff *to, |
1630 | const struct sk_buff *from) | |
1631 | { | |
9f06f87f | 1632 | #ifdef CONFIG_SKB_DECRYPTED |
41477662 JK |
1633 | to->decrypted = from->decrypted; |
1634 | #endif | |
1635 | } | |
1636 | ||
4305b541 ACM |
1637 | #ifdef NET_SKBUFF_DATA_USES_OFFSET |
1638 | static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) | |
1639 | { | |
1640 | return skb->head + skb->end; | |
1641 | } | |
ec47ea82 AD |
1642 | |
1643 | static inline unsigned int skb_end_offset(const struct sk_buff *skb) | |
1644 | { | |
1645 | return skb->end; | |
1646 | } | |
763087da ED |
1647 | |
1648 | static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset) | |
1649 | { | |
1650 | skb->end = offset; | |
1651 | } | |
4305b541 ACM |
1652 | #else |
1653 | static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) | |
1654 | { | |
1655 | return skb->end; | |
1656 | } | |
ec47ea82 AD |
1657 | |
1658 | static inline unsigned int skb_end_offset(const struct sk_buff *skb) | |
1659 | { | |
1660 | return skb->end - skb->head; | |
1661 | } | |
763087da ED |
1662 | |
1663 | static inline void skb_set_end_offset(struct sk_buff *skb, unsigned int offset) | |
1664 | { | |
1665 | skb->end = skb->head + offset; | |
1666 | } | |
4305b541 ACM |
1667 | #endif |
1668 | ||
7ab4f16f PB |
1669 | extern const struct ubuf_info_ops msg_zerocopy_ubuf_ops; |
1670 | ||
657dd5f9 PB |
1671 | struct ubuf_info *msg_zerocopy_realloc(struct sock *sk, size_t size, |
1672 | struct ubuf_info *uarg); | |
1673 | ||
1674 | void msg_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref); | |
1675 | ||
ebe73a28 DA |
1676 | int __zerocopy_sg_from_iter(struct msghdr *msg, struct sock *sk, |
1677 | struct sk_buff *skb, struct iov_iter *from, | |
1678 | size_t length); | |
657dd5f9 PB |
1679 | |
1680 | static inline int skb_zerocopy_iter_dgram(struct sk_buff *skb, | |
1681 | struct msghdr *msg, int len) | |
1682 | { | |
ebe73a28 | 1683 | return __zerocopy_sg_from_iter(msg, skb->sk, skb, &msg->msg_iter, len); |
657dd5f9 PB |
1684 | } |
1685 | ||
1686 | int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb, | |
1687 | struct msghdr *msg, int len, | |
1688 | struct ubuf_info *uarg); | |
1689 | ||
1da177e4 | 1690 | /* Internal */ |
4305b541 | 1691 | #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB))) |
1da177e4 | 1692 | |
ac45f602 PO |
1693 | static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb) |
1694 | { | |
1695 | return &skb_shinfo(skb)->hwtstamps; | |
1696 | } | |
1697 | ||
52267790 WB |
1698 | static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb) |
1699 | { | |
06b4feb3 | 1700 | bool is_zcopy = skb && skb_shinfo(skb)->flags & SKBFL_ZEROCOPY_ENABLE; |
52267790 WB |
1701 | |
1702 | return is_zcopy ? skb_uarg(skb) : NULL; | |
1703 | } | |
1704 | ||
9b65b17d TA |
1705 | static inline bool skb_zcopy_pure(const struct sk_buff *skb) |
1706 | { | |
1707 | return skb_shinfo(skb)->flags & SKBFL_PURE_ZEROCOPY; | |
1708 | } | |
1709 | ||
753f1ca4 PB |
1710 | static inline bool skb_zcopy_managed(const struct sk_buff *skb) |
1711 | { | |
1712 | return skb_shinfo(skb)->flags & SKBFL_MANAGED_FRAG_REFS; | |
1713 | } | |
1714 | ||
9b65b17d TA |
1715 | static inline bool skb_pure_zcopy_same(const struct sk_buff *skb1, |
1716 | const struct sk_buff *skb2) | |
1717 | { | |
1718 | return skb_zcopy_pure(skb1) == skb_zcopy_pure(skb2); | |
1719 | } | |
1720 | ||
8e044917 | 1721 | static inline void net_zcopy_get(struct ubuf_info *uarg) |
e76d46cf JL |
1722 | { |
1723 | refcount_inc(&uarg->refcnt); | |
1724 | } | |
1725 | ||
9ee5e5ad JL |
1726 | static inline void skb_zcopy_init(struct sk_buff *skb, struct ubuf_info *uarg) |
1727 | { | |
1728 | skb_shinfo(skb)->destructor_arg = uarg; | |
1729 | skb_shinfo(skb)->flags |= uarg->flags; | |
1730 | } | |
1731 | ||
52900d22 WB |
1732 | static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg, |
1733 | bool *have_ref) | |
52267790 WB |
1734 | { |
1735 | if (skb && uarg && !skb_zcopy(skb)) { | |
52900d22 WB |
1736 | if (unlikely(have_ref && *have_ref)) |
1737 | *have_ref = false; | |
1738 | else | |
8e044917 | 1739 | net_zcopy_get(uarg); |
9ee5e5ad | 1740 | skb_zcopy_init(skb, uarg); |
52267790 WB |
1741 | } |
1742 | } | |
1743 | ||
5cd8d46e WB |
1744 | static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val) |
1745 | { | |
1746 | skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL); | |
06b4feb3 | 1747 | skb_shinfo(skb)->flags |= SKBFL_ZEROCOPY_FRAG; |
5cd8d46e WB |
1748 | } |
1749 | ||
1750 | static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb) | |
1751 | { | |
1752 | return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL; | |
1753 | } | |
1754 | ||
1755 | static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb) | |
1756 | { | |
1757 | return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL); | |
1758 | } | |
1759 | ||
8e044917 | 1760 | static inline void net_zcopy_put(struct ubuf_info *uarg) |
59776362 JL |
1761 | { |
1762 | if (uarg) | |
7ab4f16f | 1763 | uarg->ops->complete(NULL, uarg, true); |
59776362 JL |
1764 | } |
1765 | ||
8e044917 | 1766 | static inline void net_zcopy_put_abort(struct ubuf_info *uarg, bool have_uref) |
236a6b1c JL |
1767 | { |
1768 | if (uarg) { | |
7ab4f16f | 1769 | if (uarg->ops == &msg_zerocopy_ubuf_ops) |
8c793822 | 1770 | msg_zerocopy_put_abort(uarg, have_uref); |
236a6b1c | 1771 | else if (have_uref) |
8e044917 | 1772 | net_zcopy_put(uarg); |
236a6b1c JL |
1773 | } |
1774 | } | |
1775 | ||
52267790 | 1776 | /* Release a reference on a zerocopy structure */ |
36177832 | 1777 | static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy_success) |
52267790 WB |
1778 | { |
1779 | struct ubuf_info *uarg = skb_zcopy(skb); | |
1780 | ||
1781 | if (uarg) { | |
36177832 | 1782 | if (!skb_zcopy_is_nouarg(skb)) |
7ab4f16f | 1783 | uarg->ops->complete(skb, uarg, zerocopy_success); |
0a4a060b | 1784 | |
9b65b17d | 1785 | skb_shinfo(skb)->flags &= ~SKBFL_ALL_ZEROCOPY; |
52267790 WB |
1786 | } |
1787 | } | |
1788 | ||
753f1ca4 PB |
1789 | void __skb_zcopy_downgrade_managed(struct sk_buff *skb); |
1790 | ||
1791 | static inline void skb_zcopy_downgrade_managed(struct sk_buff *skb) | |
1792 | { | |
1793 | if (unlikely(skb_zcopy_managed(skb))) | |
1794 | __skb_zcopy_downgrade_managed(skb); | |
1795 | } | |
1796 | ||
a8305bff DM |
1797 | static inline void skb_mark_not_on_list(struct sk_buff *skb) |
1798 | { | |
1799 | skb->next = NULL; | |
1800 | } | |
1801 | ||
9dde0cd3 JDB |
1802 | static inline void skb_poison_list(struct sk_buff *skb) |
1803 | { | |
1804 | #ifdef CONFIG_DEBUG_NET | |
1805 | skb->next = SKB_LIST_POISON_NEXT; | |
1806 | #endif | |
1807 | } | |
1808 | ||
dcfea72e | 1809 | /* Iterate through singly-linked GSO fragments of an skb. */ |
5eee7bd7 JD |
1810 | #define skb_list_walk_safe(first, skb, next_skb) \ |
1811 | for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \ | |
1812 | (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL) | |
dcfea72e | 1813 | |
992cba7e DM |
1814 | static inline void skb_list_del_init(struct sk_buff *skb) |
1815 | { | |
1816 | __list_del_entry(&skb->list); | |
1817 | skb_mark_not_on_list(skb); | |
1818 | } | |
1819 | ||
1da177e4 LT |
1820 | /** |
1821 | * skb_queue_empty - check if a queue is empty | |
1822 | * @list: queue head | |
1823 | * | |
1824 | * Returns true if the queue is empty, false otherwise. | |
1825 | */ | |
1826 | static inline int skb_queue_empty(const struct sk_buff_head *list) | |
1827 | { | |
fd44b93c | 1828 | return list->next == (const struct sk_buff *) list; |
1da177e4 LT |
1829 | } |
1830 | ||
d7d16a89 ED |
1831 | /** |
1832 | * skb_queue_empty_lockless - check if a queue is empty | |
1833 | * @list: queue head | |
1834 | * | |
1835 | * Returns true if the queue is empty, false otherwise. | |
1836 | * This variant can be used in lockless contexts. | |
1837 | */ | |
1838 | static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list) | |
1839 | { | |
1840 | return READ_ONCE(list->next) == (const struct sk_buff *) list; | |
1841 | } | |
1842 | ||
1843 | ||
fc7ebb21 DM |
1844 | /** |
1845 | * skb_queue_is_last - check if skb is the last entry in the queue | |
1846 | * @list: queue head | |
1847 | * @skb: buffer | |
1848 | * | |
1849 | * Returns true if @skb is the last buffer on the list. | |
1850 | */ | |
1851 | static inline bool skb_queue_is_last(const struct sk_buff_head *list, | |
1852 | const struct sk_buff *skb) | |
1853 | { | |
fd44b93c | 1854 | return skb->next == (const struct sk_buff *) list; |
fc7ebb21 DM |
1855 | } |
1856 | ||
832d11c5 IJ |
1857 | /** |
1858 | * skb_queue_is_first - check if skb is the first entry in the queue | |
1859 | * @list: queue head | |
1860 | * @skb: buffer | |
1861 | * | |
1862 | * Returns true if @skb is the first buffer on the list. | |
1863 | */ | |
1864 | static inline bool skb_queue_is_first(const struct sk_buff_head *list, | |
1865 | const struct sk_buff *skb) | |
1866 | { | |
fd44b93c | 1867 | return skb->prev == (const struct sk_buff *) list; |
832d11c5 IJ |
1868 | } |
1869 | ||
249c8b42 DM |
1870 | /** |
1871 | * skb_queue_next - return the next packet in the queue | |
1872 | * @list: queue head | |
1873 | * @skb: current buffer | |
1874 | * | |
1875 | * Return the next packet in @list after @skb. It is only valid to | |
1876 | * call this if skb_queue_is_last() evaluates to false. | |
1877 | */ | |
1878 | static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list, | |
1879 | const struct sk_buff *skb) | |
1880 | { | |
1881 | /* This BUG_ON may seem severe, but if we just return then we | |
1882 | * are going to dereference garbage. | |
1883 | */ | |
1884 | BUG_ON(skb_queue_is_last(list, skb)); | |
1885 | return skb->next; | |
1886 | } | |
1887 | ||
832d11c5 IJ |
1888 | /** |
1889 | * skb_queue_prev - return the prev packet in the queue | |
1890 | * @list: queue head | |
1891 | * @skb: current buffer | |
1892 | * | |
1893 | * Return the prev packet in @list before @skb. It is only valid to | |
1894 | * call this if skb_queue_is_first() evaluates to false. | |
1895 | */ | |
1896 | static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list, | |
1897 | const struct sk_buff *skb) | |
1898 | { | |
1899 | /* This BUG_ON may seem severe, but if we just return then we | |
1900 | * are going to dereference garbage. | |
1901 | */ | |
1902 | BUG_ON(skb_queue_is_first(list, skb)); | |
1903 | return skb->prev; | |
1904 | } | |
1905 | ||
1da177e4 LT |
1906 | /** |
1907 | * skb_get - reference buffer | |
1908 | * @skb: buffer to reference | |
1909 | * | |
1910 | * Makes another reference to a socket buffer and returns a pointer | |
1911 | * to the buffer. | |
1912 | */ | |
1913 | static inline struct sk_buff *skb_get(struct sk_buff *skb) | |
1914 | { | |
63354797 | 1915 | refcount_inc(&skb->users); |
1da177e4 LT |
1916 | return skb; |
1917 | } | |
1918 | ||
1919 | /* | |
f8821f96 | 1920 | * If users == 1, we are the only owner and can avoid redundant atomic changes. |
1da177e4 LT |
1921 | */ |
1922 | ||
1da177e4 LT |
1923 | /** |
1924 | * skb_cloned - is the buffer a clone | |
1925 | * @skb: buffer to check | |
1926 | * | |
1927 | * Returns true if the buffer was generated with skb_clone() and is | |
1928 | * one of multiple shared copies of the buffer. Cloned buffers are | |
1929 | * shared data so must not be written to under normal circumstances. | |
1930 | */ | |
1931 | static inline int skb_cloned(const struct sk_buff *skb) | |
1932 | { | |
1933 | return skb->cloned && | |
1934 | (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1; | |
1935 | } | |
1936 | ||
14bbd6a5 PS |
1937 | static inline int skb_unclone(struct sk_buff *skb, gfp_t pri) |
1938 | { | |
d0164adc | 1939 | might_sleep_if(gfpflags_allow_blocking(pri)); |
14bbd6a5 PS |
1940 | |
1941 | if (skb_cloned(skb)) | |
1942 | return pskb_expand_head(skb, 0, 0, pri); | |
1943 | ||
1944 | return 0; | |
1945 | } | |
1946 | ||
2b88cba5 ED |
1947 | /* This variant of skb_unclone() makes sure skb->truesize |
1948 | * and skb_end_offset() are not changed, whenever a new skb->head is needed. | |
1949 | * | |
1950 | * Indeed there is no guarantee that ksize(kmalloc(X)) == ksize(kmalloc(X)) | |
1951 | * when various debugging features are in place. | |
1952 | */ | |
1953 | int __skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri); | |
c4777efa ED |
1954 | static inline int skb_unclone_keeptruesize(struct sk_buff *skb, gfp_t pri) |
1955 | { | |
1956 | might_sleep_if(gfpflags_allow_blocking(pri)); | |
1957 | ||
2b88cba5 ED |
1958 | if (skb_cloned(skb)) |
1959 | return __skb_unclone_keeptruesize(skb, pri); | |
c4777efa ED |
1960 | return 0; |
1961 | } | |
1962 | ||
1da177e4 LT |
1963 | /** |
1964 | * skb_header_cloned - is the header a clone | |
1965 | * @skb: buffer to check | |
1966 | * | |
1967 | * Returns true if modifying the header part of the buffer requires | |
1968 | * the data to be copied. | |
1969 | */ | |
1970 | static inline int skb_header_cloned(const struct sk_buff *skb) | |
1971 | { | |
1972 | int dataref; | |
1973 | ||
1974 | if (!skb->cloned) | |
1975 | return 0; | |
1976 | ||
1977 | dataref = atomic_read(&skb_shinfo(skb)->dataref); | |
1978 | dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT); | |
1979 | return dataref != 1; | |
1980 | } | |
1981 | ||
9580bf2e ED |
1982 | static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri) |
1983 | { | |
1984 | might_sleep_if(gfpflags_allow_blocking(pri)); | |
1985 | ||
1986 | if (skb_header_cloned(skb)) | |
1987 | return pskb_expand_head(skb, 0, 0, pri); | |
1988 | ||
1989 | return 0; | |
1990 | } | |
1991 | ||
f4a775d1 | 1992 | /** |
9ec7ea14 JK |
1993 | * __skb_header_release() - allow clones to use the headroom |
1994 | * @skb: buffer to operate on | |
1995 | * | |
1996 | * See "DOC: dataref and headerless skbs". | |
f4a775d1 ED |
1997 | */ |
1998 | static inline void __skb_header_release(struct sk_buff *skb) | |
1999 | { | |
2000 | skb->nohdr = 1; | |
2001 | atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT)); | |
2002 | } | |
2003 | ||
2004 | ||
1da177e4 LT |
2005 | /** |
2006 | * skb_shared - is the buffer shared | |
2007 | * @skb: buffer to check | |
2008 | * | |
2009 | * Returns true if more than one person has a reference to this | |
2010 | * buffer. | |
2011 | */ | |
2012 | static inline int skb_shared(const struct sk_buff *skb) | |
2013 | { | |
63354797 | 2014 | return refcount_read(&skb->users) != 1; |
1da177e4 LT |
2015 | } |
2016 | ||
2017 | /** | |
2018 | * skb_share_check - check if buffer is shared and if so clone it | |
2019 | * @skb: buffer to check | |
2020 | * @pri: priority for memory allocation | |
2021 | * | |
2022 | * If the buffer is shared the buffer is cloned and the old copy | |
2023 | * drops a reference. A new clone with a single reference is returned. | |
2024 | * If the buffer is not shared the original buffer is returned. When | |
2025 | * being called from interrupt status or with spinlocks held pri must | |
2026 | * be GFP_ATOMIC. | |
2027 | * | |
2028 | * NULL is returned on a memory allocation failure. | |
2029 | */ | |
47061bc4 | 2030 | static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri) |
1da177e4 | 2031 | { |
d0164adc | 2032 | might_sleep_if(gfpflags_allow_blocking(pri)); |
1da177e4 LT |
2033 | if (skb_shared(skb)) { |
2034 | struct sk_buff *nskb = skb_clone(skb, pri); | |
47061bc4 ED |
2035 | |
2036 | if (likely(nskb)) | |
2037 | consume_skb(skb); | |
2038 | else | |
2039 | kfree_skb(skb); | |
1da177e4 LT |
2040 | skb = nskb; |
2041 | } | |
2042 | return skb; | |
2043 | } | |
2044 | ||
2045 | /* | |
2046 | * Copy shared buffers into a new sk_buff. We effectively do COW on | |
2047 | * packets to handle cases where we have a local reader and forward | |
2048 | * and a couple of other messy ones. The normal one is tcpdumping | |
8527ca77 | 2049 | * a packet that's being forwarded. |
1da177e4 LT |
2050 | */ |
2051 | ||
2052 | /** | |
2053 | * skb_unshare - make a copy of a shared buffer | |
2054 | * @skb: buffer to check | |
2055 | * @pri: priority for memory allocation | |
2056 | * | |
2057 | * If the socket buffer is a clone then this function creates a new | |
2058 | * copy of the data, drops a reference count on the old copy and returns | |
2059 | * the new copy with the reference count at 1. If the buffer is not a clone | |
2060 | * the original buffer is returned. When called with a spinlock held or | |
2061 | * from interrupt state @pri must be %GFP_ATOMIC | |
2062 | * | |
2063 | * %NULL is returned on a memory allocation failure. | |
2064 | */ | |
e2bf521d | 2065 | static inline struct sk_buff *skb_unshare(struct sk_buff *skb, |
dd0fc66f | 2066 | gfp_t pri) |
1da177e4 | 2067 | { |
d0164adc | 2068 | might_sleep_if(gfpflags_allow_blocking(pri)); |
1da177e4 LT |
2069 | if (skb_cloned(skb)) { |
2070 | struct sk_buff *nskb = skb_copy(skb, pri); | |
31eff81e AA |
2071 | |
2072 | /* Free our shared copy */ | |
2073 | if (likely(nskb)) | |
2074 | consume_skb(skb); | |
2075 | else | |
2076 | kfree_skb(skb); | |
1da177e4 LT |
2077 | skb = nskb; |
2078 | } | |
2079 | return skb; | |
2080 | } | |
2081 | ||
2082 | /** | |
1a5778aa | 2083 | * skb_peek - peek at the head of an &sk_buff_head |
1da177e4 LT |
2084 | * @list_: list to peek at |
2085 | * | |
2086 | * Peek an &sk_buff. Unlike most other operations you _MUST_ | |
2087 | * be careful with this one. A peek leaves the buffer on the | |
2088 | * list and someone else may run off with it. You must hold | |
2089 | * the appropriate locks or have a private queue to do this. | |
2090 | * | |
2091 | * Returns %NULL for an empty list or a pointer to the head element. | |
2092 | * The reference count is not incremented and the reference is therefore | |
2093 | * volatile. Use with caution. | |
2094 | */ | |
05bdd2f1 | 2095 | static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_) |
1da177e4 | 2096 | { |
18d07000 ED |
2097 | struct sk_buff *skb = list_->next; |
2098 | ||
2099 | if (skb == (struct sk_buff *)list_) | |
2100 | skb = NULL; | |
2101 | return skb; | |
1da177e4 LT |
2102 | } |
2103 | ||
8b69bd7d DM |
2104 | /** |
2105 | * __skb_peek - peek at the head of a non-empty &sk_buff_head | |
2106 | * @list_: list to peek at | |
2107 | * | |
2108 | * Like skb_peek(), but the caller knows that the list is not empty. | |
2109 | */ | |
2110 | static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_) | |
2111 | { | |
2112 | return list_->next; | |
2113 | } | |
2114 | ||
da5ef6e5 PE |
2115 | /** |
2116 | * skb_peek_next - peek skb following the given one from a queue | |
2117 | * @skb: skb to start from | |
2118 | * @list_: list to peek at | |
2119 | * | |
2120 | * Returns %NULL when the end of the list is met or a pointer to the | |
2121 | * next element. The reference count is not incremented and the | |
2122 | * reference is therefore volatile. Use with caution. | |
2123 | */ | |
2124 | static inline struct sk_buff *skb_peek_next(struct sk_buff *skb, | |
2125 | const struct sk_buff_head *list_) | |
2126 | { | |
2127 | struct sk_buff *next = skb->next; | |
18d07000 | 2128 | |
da5ef6e5 PE |
2129 | if (next == (struct sk_buff *)list_) |
2130 | next = NULL; | |
2131 | return next; | |
2132 | } | |
2133 | ||
1da177e4 | 2134 | /** |
1a5778aa | 2135 | * skb_peek_tail - peek at the tail of an &sk_buff_head |
1da177e4 LT |
2136 | * @list_: list to peek at |
2137 | * | |
2138 | * Peek an &sk_buff. Unlike most other operations you _MUST_ | |
2139 | * be careful with this one. A peek leaves the buffer on the | |
2140 | * list and someone else may run off with it. You must hold | |
2141 | * the appropriate locks or have a private queue to do this. | |
2142 | * | |
2143 | * Returns %NULL for an empty list or a pointer to the tail element. | |
2144 | * The reference count is not incremented and the reference is therefore | |
2145 | * volatile. Use with caution. | |
2146 | */ | |
05bdd2f1 | 2147 | static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_) |
1da177e4 | 2148 | { |
f8cc62ca | 2149 | struct sk_buff *skb = READ_ONCE(list_->prev); |
18d07000 ED |
2150 | |
2151 | if (skb == (struct sk_buff *)list_) | |
2152 | skb = NULL; | |
2153 | return skb; | |
2154 | ||
1da177e4 LT |
2155 | } |
2156 | ||
2157 | /** | |
2158 | * skb_queue_len - get queue length | |
2159 | * @list_: list to measure | |
2160 | * | |
2161 | * Return the length of an &sk_buff queue. | |
2162 | */ | |
2163 | static inline __u32 skb_queue_len(const struct sk_buff_head *list_) | |
2164 | { | |
2165 | return list_->qlen; | |
2166 | } | |
2167 | ||
86b18aaa QC |
2168 | /** |
2169 | * skb_queue_len_lockless - get queue length | |
2170 | * @list_: list to measure | |
2171 | * | |
2172 | * Return the length of an &sk_buff queue. | |
2173 | * This variant can be used in lockless contexts. | |
2174 | */ | |
2175 | static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_) | |
2176 | { | |
2177 | return READ_ONCE(list_->qlen); | |
2178 | } | |
2179 | ||
67fed459 DM |
2180 | /** |
2181 | * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head | |
2182 | * @list: queue to initialize | |
2183 | * | |
2184 | * This initializes only the list and queue length aspects of | |
2185 | * an sk_buff_head object. This allows to initialize the list | |
2186 | * aspects of an sk_buff_head without reinitializing things like | |
2187 | * the spinlock. It can also be used for on-stack sk_buff_head | |
2188 | * objects where the spinlock is known to not be used. | |
2189 | */ | |
2190 | static inline void __skb_queue_head_init(struct sk_buff_head *list) | |
2191 | { | |
2192 | list->prev = list->next = (struct sk_buff *)list; | |
2193 | list->qlen = 0; | |
2194 | } | |
2195 | ||
76f10ad0 AV |
2196 | /* |
2197 | * This function creates a split out lock class for each invocation; | |
2198 | * this is needed for now since a whole lot of users of the skb-queue | |
2199 | * infrastructure in drivers have different locking usage (in hardirq) | |
2200 | * than the networking core (in softirq only). In the long run either the | |
2201 | * network layer or drivers should need annotation to consolidate the | |
2202 | * main types of usage into 3 classes. | |
2203 | */ | |
1da177e4 LT |
2204 | static inline void skb_queue_head_init(struct sk_buff_head *list) |
2205 | { | |
2206 | spin_lock_init(&list->lock); | |
67fed459 | 2207 | __skb_queue_head_init(list); |
1da177e4 LT |
2208 | } |
2209 | ||
c2ecba71 PE |
2210 | static inline void skb_queue_head_init_class(struct sk_buff_head *list, |
2211 | struct lock_class_key *class) | |
2212 | { | |
2213 | skb_queue_head_init(list); | |
2214 | lockdep_set_class(&list->lock, class); | |
2215 | } | |
2216 | ||
1da177e4 | 2217 | /* |
bf299275 | 2218 | * Insert an sk_buff on a list. |
1da177e4 LT |
2219 | * |
2220 | * The "__skb_xxxx()" functions are the non-atomic ones that | |
2221 | * can only be called with interrupts disabled. | |
2222 | */ | |
bf299275 GR |
2223 | static inline void __skb_insert(struct sk_buff *newsk, |
2224 | struct sk_buff *prev, struct sk_buff *next, | |
2225 | struct sk_buff_head *list) | |
2226 | { | |
f8cc62ca ED |
2227 | /* See skb_queue_empty_lockless() and skb_peek_tail() |
2228 | * for the opposite READ_ONCE() | |
2229 | */ | |
d7d16a89 ED |
2230 | WRITE_ONCE(newsk->next, next); |
2231 | WRITE_ONCE(newsk->prev, prev); | |
1a2fb220 KC |
2232 | WRITE_ONCE(((struct sk_buff_list *)next)->prev, newsk); |
2233 | WRITE_ONCE(((struct sk_buff_list *)prev)->next, newsk); | |
04f08eb4 | 2234 | WRITE_ONCE(list->qlen, list->qlen + 1); |
bf299275 | 2235 | } |
1da177e4 | 2236 | |
67fed459 DM |
2237 | static inline void __skb_queue_splice(const struct sk_buff_head *list, |
2238 | struct sk_buff *prev, | |
2239 | struct sk_buff *next) | |
2240 | { | |
2241 | struct sk_buff *first = list->next; | |
2242 | struct sk_buff *last = list->prev; | |
2243 | ||
d7d16a89 ED |
2244 | WRITE_ONCE(first->prev, prev); |
2245 | WRITE_ONCE(prev->next, first); | |
67fed459 | 2246 | |
d7d16a89 ED |
2247 | WRITE_ONCE(last->next, next); |
2248 | WRITE_ONCE(next->prev, last); | |
67fed459 DM |
2249 | } |
2250 | ||
2251 | /** | |
2252 | * skb_queue_splice - join two skb lists, this is designed for stacks | |
2253 | * @list: the new list to add | |
2254 | * @head: the place to add it in the first list | |
2255 | */ | |
2256 | static inline void skb_queue_splice(const struct sk_buff_head *list, | |
2257 | struct sk_buff_head *head) | |
2258 | { | |
2259 | if (!skb_queue_empty(list)) { | |
2260 | __skb_queue_splice(list, (struct sk_buff *) head, head->next); | |
1d4a31dd | 2261 | head->qlen += list->qlen; |
67fed459 DM |
2262 | } |
2263 | } | |
2264 | ||
2265 | /** | |
d9619496 | 2266 | * skb_queue_splice_init - join two skb lists and reinitialise the emptied list |
67fed459 DM |
2267 | * @list: the new list to add |
2268 | * @head: the place to add it in the first list | |
2269 | * | |
2270 | * The list at @list is reinitialised | |
2271 | */ | |
2272 | static inline void skb_queue_splice_init(struct sk_buff_head *list, | |
2273 | struct sk_buff_head *head) | |
2274 | { | |
2275 | if (!skb_queue_empty(list)) { | |
2276 | __skb_queue_splice(list, (struct sk_buff *) head, head->next); | |
1d4a31dd | 2277 | head->qlen += list->qlen; |
67fed459 DM |
2278 | __skb_queue_head_init(list); |
2279 | } | |
2280 | } | |
2281 | ||
2282 | /** | |
2283 | * skb_queue_splice_tail - join two skb lists, each list being a queue | |
2284 | * @list: the new list to add | |
2285 | * @head: the place to add it in the first list | |
2286 | */ | |
2287 | static inline void skb_queue_splice_tail(const struct sk_buff_head *list, | |
2288 | struct sk_buff_head *head) | |
2289 | { | |
2290 | if (!skb_queue_empty(list)) { | |
2291 | __skb_queue_splice(list, head->prev, (struct sk_buff *) head); | |
1d4a31dd | 2292 | head->qlen += list->qlen; |
67fed459 DM |
2293 | } |
2294 | } | |
2295 | ||
2296 | /** | |
d9619496 | 2297 | * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list |
67fed459 DM |
2298 | * @list: the new list to add |
2299 | * @head: the place to add it in the first list | |
2300 | * | |
2301 | * Each of the lists is a queue. | |
2302 | * The list at @list is reinitialised | |
2303 | */ | |
2304 | static inline void skb_queue_splice_tail_init(struct sk_buff_head *list, | |
2305 | struct sk_buff_head *head) | |
2306 | { | |
2307 | if (!skb_queue_empty(list)) { | |
2308 | __skb_queue_splice(list, head->prev, (struct sk_buff *) head); | |
1d4a31dd | 2309 | head->qlen += list->qlen; |
67fed459 DM |
2310 | __skb_queue_head_init(list); |
2311 | } | |
2312 | } | |
2313 | ||
1da177e4 | 2314 | /** |
300ce174 | 2315 | * __skb_queue_after - queue a buffer at the list head |
1da177e4 | 2316 | * @list: list to use |
300ce174 | 2317 | * @prev: place after this buffer |
1da177e4 LT |
2318 | * @newsk: buffer to queue |
2319 | * | |
300ce174 | 2320 | * Queue a buffer int the middle of a list. This function takes no locks |
1da177e4 LT |
2321 | * and you must therefore hold required locks before calling it. |
2322 | * | |
2323 | * A buffer cannot be placed on two lists at the same time. | |
2324 | */ | |
300ce174 SH |
2325 | static inline void __skb_queue_after(struct sk_buff_head *list, |
2326 | struct sk_buff *prev, | |
2327 | struct sk_buff *newsk) | |
1da177e4 | 2328 | { |
1a2fb220 | 2329 | __skb_insert(newsk, prev, ((struct sk_buff_list *)prev)->next, list); |
1da177e4 LT |
2330 | } |
2331 | ||
7965bd4d JP |
2332 | void skb_append(struct sk_buff *old, struct sk_buff *newsk, |
2333 | struct sk_buff_head *list); | |
7de6c033 | 2334 | |
f5572855 GR |
2335 | static inline void __skb_queue_before(struct sk_buff_head *list, |
2336 | struct sk_buff *next, | |
2337 | struct sk_buff *newsk) | |
2338 | { | |
1a2fb220 | 2339 | __skb_insert(newsk, ((struct sk_buff_list *)next)->prev, next, list); |
f5572855 GR |
2340 | } |
2341 | ||
300ce174 SH |
2342 | /** |
2343 | * __skb_queue_head - queue a buffer at the list head | |
2344 | * @list: list to use | |
2345 | * @newsk: buffer to queue | |
2346 | * | |
2347 | * Queue a buffer at the start of a list. This function takes no locks | |
2348 | * and you must therefore hold required locks before calling it. | |
2349 | * | |
2350 | * A buffer cannot be placed on two lists at the same time. | |
2351 | */ | |
300ce174 SH |
2352 | static inline void __skb_queue_head(struct sk_buff_head *list, |
2353 | struct sk_buff *newsk) | |
2354 | { | |
2355 | __skb_queue_after(list, (struct sk_buff *)list, newsk); | |
2356 | } | |
4ea7b0cf | 2357 | void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk); |
300ce174 | 2358 | |
1da177e4 LT |
2359 | /** |
2360 | * __skb_queue_tail - queue a buffer at the list tail | |
2361 | * @list: list to use | |
2362 | * @newsk: buffer to queue | |
2363 | * | |
2364 | * Queue a buffer at the end of a list. This function takes no locks | |
2365 | * and you must therefore hold required locks before calling it. | |
2366 | * | |
2367 | * A buffer cannot be placed on two lists at the same time. | |
2368 | */ | |
1da177e4 LT |
2369 | static inline void __skb_queue_tail(struct sk_buff_head *list, |
2370 | struct sk_buff *newsk) | |
2371 | { | |
f5572855 | 2372 | __skb_queue_before(list, (struct sk_buff *)list, newsk); |
1da177e4 | 2373 | } |
4ea7b0cf | 2374 | void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk); |
1da177e4 | 2375 | |
1da177e4 LT |
2376 | /* |
2377 | * remove sk_buff from list. _Must_ be called atomically, and with | |
2378 | * the list known.. | |
2379 | */ | |
7965bd4d | 2380 | void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list); |
1da177e4 LT |
2381 | static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list) |
2382 | { | |
2383 | struct sk_buff *next, *prev; | |
2384 | ||
86b18aaa | 2385 | WRITE_ONCE(list->qlen, list->qlen - 1); |
1da177e4 LT |
2386 | next = skb->next; |
2387 | prev = skb->prev; | |
2388 | skb->next = skb->prev = NULL; | |
d7d16a89 ED |
2389 | WRITE_ONCE(next->prev, prev); |
2390 | WRITE_ONCE(prev->next, next); | |
1da177e4 LT |
2391 | } |
2392 | ||
f525c06d GR |
2393 | /** |
2394 | * __skb_dequeue - remove from the head of the queue | |
2395 | * @list: list to dequeue from | |
2396 | * | |
2397 | * Remove the head of the list. This function does not take any locks | |
2398 | * so must be used with appropriate locks held only. The head item is | |
2399 | * returned or %NULL if the list is empty. | |
2400 | */ | |
f525c06d GR |
2401 | static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list) |
2402 | { | |
2403 | struct sk_buff *skb = skb_peek(list); | |
2404 | if (skb) | |
2405 | __skb_unlink(skb, list); | |
2406 | return skb; | |
2407 | } | |
4ea7b0cf | 2408 | struct sk_buff *skb_dequeue(struct sk_buff_head *list); |
1da177e4 LT |
2409 | |
2410 | /** | |
2411 | * __skb_dequeue_tail - remove from the tail of the queue | |
2412 | * @list: list to dequeue from | |
2413 | * | |
2414 | * Remove the tail of the list. This function does not take any locks | |
2415 | * so must be used with appropriate locks held only. The tail item is | |
2416 | * returned or %NULL if the list is empty. | |
2417 | */ | |
1da177e4 LT |
2418 | static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list) |
2419 | { | |
2420 | struct sk_buff *skb = skb_peek_tail(list); | |
2421 | if (skb) | |
2422 | __skb_unlink(skb, list); | |
2423 | return skb; | |
2424 | } | |
4ea7b0cf | 2425 | struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list); |
1da177e4 LT |
2426 | |
2427 | ||
bdcc0924 | 2428 | static inline bool skb_is_nonlinear(const struct sk_buff *skb) |
1da177e4 LT |
2429 | { |
2430 | return skb->data_len; | |
2431 | } | |
2432 | ||
2433 | static inline unsigned int skb_headlen(const struct sk_buff *skb) | |
2434 | { | |
2435 | return skb->len - skb->data_len; | |
2436 | } | |
2437 | ||
3ece7826 | 2438 | static inline unsigned int __skb_pagelen(const struct sk_buff *skb) |
1da177e4 | 2439 | { |
c72d8cda | 2440 | unsigned int i, len = 0; |
1da177e4 | 2441 | |
c72d8cda | 2442 | for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--) |
9e903e08 | 2443 | len += skb_frag_size(&skb_shinfo(skb)->frags[i]); |
3ece7826 WB |
2444 | return len; |
2445 | } | |
2446 | ||
2447 | static inline unsigned int skb_pagelen(const struct sk_buff *skb) | |
2448 | { | |
2449 | return skb_headlen(skb) + __skb_pagelen(skb); | |
1da177e4 LT |
2450 | } |
2451 | ||
21d2e673 MA |
2452 | static inline void skb_frag_fill_netmem_desc(skb_frag_t *frag, |
2453 | netmem_ref netmem, int off, | |
2454 | int size) | |
2455 | { | |
2456 | frag->netmem = netmem; | |
2457 | frag->offset = off; | |
2458 | skb_frag_size_set(frag, size); | |
2459 | } | |
2460 | ||
b51f4113 YL |
2461 | static inline void skb_frag_fill_page_desc(skb_frag_t *frag, |
2462 | struct page *page, | |
2463 | int off, int size) | |
2464 | { | |
21d2e673 MA |
2465 | skb_frag_fill_netmem_desc(frag, page_to_netmem(page), off, size); |
2466 | } | |
2467 | ||
2468 | static inline void __skb_fill_netmem_desc_noacc(struct skb_shared_info *shinfo, | |
2469 | int i, netmem_ref netmem, | |
2470 | int off, int size) | |
2471 | { | |
2472 | skb_frag_t *frag = &shinfo->frags[i]; | |
2473 | ||
2474 | skb_frag_fill_netmem_desc(frag, netmem, off, size); | |
b51f4113 YL |
2475 | } |
2476 | ||
84ce071e PB |
2477 | static inline void __skb_fill_page_desc_noacc(struct skb_shared_info *shinfo, |
2478 | int i, struct page *page, | |
2479 | int off, int size) | |
2480 | { | |
21d2e673 MA |
2481 | __skb_fill_netmem_desc_noacc(shinfo, i, page_to_netmem(page), off, |
2482 | size); | |
84ce071e PB |
2483 | } |
2484 | ||
ede57d58 RG |
2485 | /** |
2486 | * skb_len_add - adds a number to len fields of skb | |
2487 | * @skb: buffer to add len to | |
2488 | * @delta: number of bytes to add | |
2489 | */ | |
2490 | static inline void skb_len_add(struct sk_buff *skb, int delta) | |
2491 | { | |
2492 | skb->len += delta; | |
2493 | skb->data_len += delta; | |
2494 | skb->truesize += delta; | |
2495 | } | |
2496 | ||
131ea667 | 2497 | /** |
21d2e673 | 2498 | * __skb_fill_netmem_desc - initialise a fragment in an skb |
131ea667 | 2499 | * @skb: buffer containing fragment to be initialised |
21d2e673 MA |
2500 | * @i: fragment index to initialise |
2501 | * @netmem: the netmem to use for this fragment | |
131ea667 IC |
2502 | * @off: the offset to the data with @page |
2503 | * @size: the length of the data | |
2504 | * | |
2505 | * Initialises the @i'th fragment of @skb to point to &size bytes at | |
2506 | * offset @off within @page. | |
2507 | * | |
2508 | * Does not take any additional reference on the fragment. | |
2509 | */ | |
21d2e673 MA |
2510 | static inline void __skb_fill_netmem_desc(struct sk_buff *skb, int i, |
2511 | netmem_ref netmem, int off, int size) | |
1da177e4 | 2512 | { |
21d2e673 MA |
2513 | struct page *page = netmem_to_page(netmem); |
2514 | ||
2515 | __skb_fill_netmem_desc_noacc(skb_shinfo(skb), i, netmem, off, size); | |
8b334851 YL |
2516 | |
2517 | /* Propagate page pfmemalloc to the skb if we can. The problem is | |
2518 | * that not all callers have unique ownership of the page but rely | |
2519 | * on page_is_pfmemalloc doing the right thing(tm). | |
2520 | */ | |
cca7af38 | 2521 | page = compound_head(page); |
2f064f34 | 2522 | if (page_is_pfmemalloc(page)) |
21d2e673 MA |
2523 | skb->pfmemalloc = true; |
2524 | } | |
2525 | ||
2526 | static inline void __skb_fill_page_desc(struct sk_buff *skb, int i, | |
2527 | struct page *page, int off, int size) | |
2528 | { | |
2529 | __skb_fill_netmem_desc(skb, i, page_to_netmem(page), off, size); | |
2530 | } | |
2531 | ||
2532 | static inline void skb_fill_netmem_desc(struct sk_buff *skb, int i, | |
2533 | netmem_ref netmem, int off, int size) | |
2534 | { | |
2535 | __skb_fill_netmem_desc(skb, i, netmem, off, size); | |
2536 | skb_shinfo(skb)->nr_frags = i + 1; | |
131ea667 IC |
2537 | } |
2538 | ||
2539 | /** | |
2540 | * skb_fill_page_desc - initialise a paged fragment in an skb | |
2541 | * @skb: buffer containing fragment to be initialised | |
2542 | * @i: paged fragment index to initialise | |
2543 | * @page: the page to use for this fragment | |
2544 | * @off: the offset to the data with @page | |
2545 | * @size: the length of the data | |
2546 | * | |
2547 | * As per __skb_fill_page_desc() -- initialises the @i'th fragment of | |
bc32383c | 2548 | * @skb to point to @size bytes at offset @off within @page. In |
131ea667 IC |
2549 | * addition updates @skb such that @i is the last fragment. |
2550 | * | |
2551 | * Does not take any additional reference on the fragment. | |
2552 | */ | |
2553 | static inline void skb_fill_page_desc(struct sk_buff *skb, int i, | |
2554 | struct page *page, int off, int size) | |
2555 | { | |
21d2e673 | 2556 | skb_fill_netmem_desc(skb, i, page_to_netmem(page), off, size); |
1da177e4 LT |
2557 | } |
2558 | ||
32614006 ED |
2559 | /** |
2560 | * skb_fill_page_desc_noacc - initialise a paged fragment in an skb | |
2561 | * @skb: buffer containing fragment to be initialised | |
2562 | * @i: paged fragment index to initialise | |
2563 | * @page: the page to use for this fragment | |
2564 | * @off: the offset to the data with @page | |
2565 | * @size: the length of the data | |
2566 | * | |
2567 | * Variant of skb_fill_page_desc() which does not deal with | |
2568 | * pfmemalloc, if page is not owned by us. | |
2569 | */ | |
2570 | static inline void skb_fill_page_desc_noacc(struct sk_buff *skb, int i, | |
2571 | struct page *page, int off, | |
2572 | int size) | |
2573 | { | |
2574 | struct skb_shared_info *shinfo = skb_shinfo(skb); | |
2575 | ||
2576 | __skb_fill_page_desc_noacc(shinfo, i, page, off, size); | |
2577 | shinfo->nr_frags = i + 1; | |
2578 | } | |
2579 | ||
21d2e673 MA |
2580 | void skb_add_rx_frag_netmem(struct sk_buff *skb, int i, netmem_ref netmem, |
2581 | int off, int size, unsigned int truesize); | |
2582 | ||
2583 | static inline void skb_add_rx_frag(struct sk_buff *skb, int i, | |
2584 | struct page *page, int off, int size, | |
2585 | unsigned int truesize) | |
2586 | { | |
2587 | skb_add_rx_frag_netmem(skb, i, page_to_netmem(page), off, size, | |
2588 | truesize); | |
2589 | } | |
654bed16 | 2590 | |
f8e617e1 JW |
2591 | void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size, |
2592 | unsigned int truesize); | |
2593 | ||
1da177e4 LT |
2594 | #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb)) |
2595 | ||
27a884dc ACM |
2596 | #ifdef NET_SKBUFF_DATA_USES_OFFSET |
2597 | static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) | |
2598 | { | |
2599 | return skb->head + skb->tail; | |
2600 | } | |
2601 | ||
2602 | static inline void skb_reset_tail_pointer(struct sk_buff *skb) | |
2603 | { | |
2604 | skb->tail = skb->data - skb->head; | |
2605 | } | |
2606 | ||
2607 | static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) | |
2608 | { | |
2609 | skb_reset_tail_pointer(skb); | |
2610 | skb->tail += offset; | |
2611 | } | |
7cc46190 | 2612 | |
27a884dc ACM |
2613 | #else /* NET_SKBUFF_DATA_USES_OFFSET */ |
2614 | static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) | |
2615 | { | |
2616 | return skb->tail; | |
2617 | } | |
2618 | ||
2619 | static inline void skb_reset_tail_pointer(struct sk_buff *skb) | |
2620 | { | |
2621 | skb->tail = skb->data; | |
2622 | } | |
2623 | ||
2624 | static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) | |
2625 | { | |
2626 | skb->tail = skb->data + offset; | |
2627 | } | |
4305b541 | 2628 | |
27a884dc ACM |
2629 | #endif /* NET_SKBUFF_DATA_USES_OFFSET */ |
2630 | ||
fd189422 ZS |
2631 | static inline void skb_assert_len(struct sk_buff *skb) |
2632 | { | |
2633 | #ifdef CONFIG_DEBUG_NET | |
2634 | if (WARN_ONCE(!skb->len, "%s\n", __func__)) | |
2635 | DO_ONCE_LITE(skb_dump, KERN_ERR, skb, false); | |
2636 | #endif /* CONFIG_DEBUG_NET */ | |
2637 | } | |
2638 | ||
1da177e4 LT |
2639 | /* |
2640 | * Add data to an sk_buff | |
2641 | */ | |
4df864c1 JB |
2642 | void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len); |
2643 | void *skb_put(struct sk_buff *skb, unsigned int len); | |
2644 | static inline void *__skb_put(struct sk_buff *skb, unsigned int len) | |
1da177e4 | 2645 | { |
4df864c1 | 2646 | void *tmp = skb_tail_pointer(skb); |
1da177e4 LT |
2647 | SKB_LINEAR_ASSERT(skb); |
2648 | skb->tail += len; | |
2649 | skb->len += len; | |
2650 | return tmp; | |
2651 | } | |
2652 | ||
de77b966 | 2653 | static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len) |
2654 | { | |
2655 | void *tmp = __skb_put(skb, len); | |
2656 | ||
2657 | memset(tmp, 0, len); | |
2658 | return tmp; | |
2659 | } | |
2660 | ||
2661 | static inline void *__skb_put_data(struct sk_buff *skb, const void *data, | |
2662 | unsigned int len) | |
2663 | { | |
2664 | void *tmp = __skb_put(skb, len); | |
2665 | ||
2666 | memcpy(tmp, data, len); | |
2667 | return tmp; | |
2668 | } | |
2669 | ||
2670 | static inline void __skb_put_u8(struct sk_buff *skb, u8 val) | |
2671 | { | |
2672 | *(u8 *)__skb_put(skb, 1) = val; | |
2673 | } | |
2674 | ||
83ad357d | 2675 | static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len) |
e45a79da | 2676 | { |
83ad357d | 2677 | void *tmp = skb_put(skb, len); |
e45a79da JB |
2678 | |
2679 | memset(tmp, 0, len); | |
2680 | ||
2681 | return tmp; | |
2682 | } | |
2683 | ||
59ae1d12 JB |
2684 | static inline void *skb_put_data(struct sk_buff *skb, const void *data, |
2685 | unsigned int len) | |
2686 | { | |
2687 | void *tmp = skb_put(skb, len); | |
2688 | ||
2689 | memcpy(tmp, data, len); | |
2690 | ||
2691 | return tmp; | |
2692 | } | |
2693 | ||
634fef61 JB |
2694 | static inline void skb_put_u8(struct sk_buff *skb, u8 val) |
2695 | { | |
2696 | *(u8 *)skb_put(skb, 1) = val; | |
2697 | } | |
2698 | ||
d58ff351 JB |
2699 | void *skb_push(struct sk_buff *skb, unsigned int len); |
2700 | static inline void *__skb_push(struct sk_buff *skb, unsigned int len) | |
1da177e4 | 2701 | { |
219eee9c FW |
2702 | DEBUG_NET_WARN_ON_ONCE(len > INT_MAX); |
2703 | ||
1da177e4 LT |
2704 | skb->data -= len; |
2705 | skb->len += len; | |
2706 | return skb->data; | |
2707 | } | |
2708 | ||
af72868b JB |
2709 | void *skb_pull(struct sk_buff *skb, unsigned int len); |
2710 | static inline void *__skb_pull(struct sk_buff *skb, unsigned int len) | |
1da177e4 | 2711 | { |
219eee9c FW |
2712 | DEBUG_NET_WARN_ON_ONCE(len > INT_MAX); |
2713 | ||
1da177e4 | 2714 | skb->len -= len; |
22296a5c ED |
2715 | if (unlikely(skb->len < skb->data_len)) { |
2716 | #if defined(CONFIG_DEBUG_NET) | |
2717 | skb->len += len; | |
2718 | pr_err("__skb_pull(len=%u)\n", len); | |
2719 | skb_dump(KERN_ERR, skb, false); | |
2720 | #endif | |
2721 | BUG(); | |
2722 | } | |
1da177e4 LT |
2723 | return skb->data += len; |
2724 | } | |
2725 | ||
af72868b | 2726 | static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len) |
47d29646 DM |
2727 | { |
2728 | return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len); | |
2729 | } | |
2730 | ||
13244ccc LAD |
2731 | void *skb_pull_data(struct sk_buff *skb, size_t len); |
2732 | ||
af72868b | 2733 | void *__pskb_pull_tail(struct sk_buff *skb, int delta); |
1da177e4 | 2734 | |
1fb2d415 ED |
2735 | static inline enum skb_drop_reason |
2736 | pskb_may_pull_reason(struct sk_buff *skb, unsigned int len) | |
1da177e4 | 2737 | { |
219eee9c FW |
2738 | DEBUG_NET_WARN_ON_ONCE(len > INT_MAX); |
2739 | ||
1da177e4 | 2740 | if (likely(len <= skb_headlen(skb))) |
1fb2d415 ED |
2741 | return SKB_NOT_DROPPED_YET; |
2742 | ||
1da177e4 | 2743 | if (unlikely(len > skb->len)) |
1fb2d415 ED |
2744 | return SKB_DROP_REASON_PKT_TOO_SMALL; |
2745 | ||
2746 | if (unlikely(!__pskb_pull_tail(skb, len - skb_headlen(skb)))) | |
2747 | return SKB_DROP_REASON_NOMEM; | |
2748 | ||
2749 | return SKB_NOT_DROPPED_YET; | |
2750 | } | |
2751 | ||
2752 | static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len) | |
2753 | { | |
2754 | return pskb_may_pull_reason(skb, len) == SKB_NOT_DROPPED_YET; | |
1da177e4 LT |
2755 | } |
2756 | ||
d427c899 RG |
2757 | static inline void *pskb_pull(struct sk_buff *skb, unsigned int len) |
2758 | { | |
2759 | if (!pskb_may_pull(skb, len)) | |
2760 | return NULL; | |
2761 | ||
2762 | skb->len -= len; | |
2763 | return skb->data += len; | |
2764 | } | |
2765 | ||
c8c8b127 ED |
2766 | void skb_condense(struct sk_buff *skb); |
2767 | ||
1da177e4 LT |
2768 | /** |
2769 | * skb_headroom - bytes at buffer head | |
2770 | * @skb: buffer to check | |
2771 | * | |
2772 | * Return the number of bytes of free space at the head of an &sk_buff. | |
2773 | */ | |
c2636b4d | 2774 | static inline unsigned int skb_headroom(const struct sk_buff *skb) |
1da177e4 LT |
2775 | { |
2776 | return skb->data - skb->head; | |
2777 | } | |
2778 | ||
2779 | /** | |
2780 | * skb_tailroom - bytes at buffer end | |
2781 | * @skb: buffer to check | |
2782 | * | |
2783 | * Return the number of bytes of free space at the tail of an sk_buff | |
2784 | */ | |
2785 | static inline int skb_tailroom(const struct sk_buff *skb) | |
2786 | { | |
4305b541 | 2787 | return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail; |
1da177e4 LT |
2788 | } |
2789 | ||
a21d4572 ED |
2790 | /** |
2791 | * skb_availroom - bytes at buffer end | |
2792 | * @skb: buffer to check | |
2793 | * | |
2794 | * Return the number of bytes of free space at the tail of an sk_buff | |
2795 | * allocated by sk_stream_alloc() | |
2796 | */ | |
2797 | static inline int skb_availroom(const struct sk_buff *skb) | |
2798 | { | |
16fad69c ED |
2799 | if (skb_is_nonlinear(skb)) |
2800 | return 0; | |
2801 | ||
2802 | return skb->end - skb->tail - skb->reserved_tailroom; | |
a21d4572 ED |
2803 | } |
2804 | ||
1da177e4 LT |
2805 | /** |
2806 | * skb_reserve - adjust headroom | |
2807 | * @skb: buffer to alter | |
2808 | * @len: bytes to move | |
2809 | * | |
2810 | * Increase the headroom of an empty &sk_buff by reducing the tail | |
2811 | * room. This is only allowed for an empty buffer. | |
2812 | */ | |
8243126c | 2813 | static inline void skb_reserve(struct sk_buff *skb, int len) |
1da177e4 LT |
2814 | { |
2815 | skb->data += len; | |
2816 | skb->tail += len; | |
2817 | } | |
2818 | ||
1837b2e2 BP |
2819 | /** |
2820 | * skb_tailroom_reserve - adjust reserved_tailroom | |
2821 | * @skb: buffer to alter | |
2822 | * @mtu: maximum amount of headlen permitted | |
2823 | * @needed_tailroom: minimum amount of reserved_tailroom | |
2824 | * | |
2825 | * Set reserved_tailroom so that headlen can be as large as possible but | |
2826 | * not larger than mtu and tailroom cannot be smaller than | |
2827 | * needed_tailroom. | |
2828 | * The required headroom should already have been reserved before using | |
2829 | * this function. | |
2830 | */ | |
2831 | static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu, | |
2832 | unsigned int needed_tailroom) | |
2833 | { | |
2834 | SKB_LINEAR_ASSERT(skb); | |
2835 | if (mtu < skb_tailroom(skb) - needed_tailroom) | |
2836 | /* use at most mtu */ | |
2837 | skb->reserved_tailroom = skb_tailroom(skb) - mtu; | |
2838 | else | |
2839 | /* use up to all available space */ | |
2840 | skb->reserved_tailroom = needed_tailroom; | |
2841 | } | |
2842 | ||
8bce6d7d TH |
2843 | #define ENCAP_TYPE_ETHER 0 |
2844 | #define ENCAP_TYPE_IPPROTO 1 | |
2845 | ||
2846 | static inline void skb_set_inner_protocol(struct sk_buff *skb, | |
2847 | __be16 protocol) | |
2848 | { | |
2849 | skb->inner_protocol = protocol; | |
2850 | skb->inner_protocol_type = ENCAP_TYPE_ETHER; | |
2851 | } | |
2852 | ||
2853 | static inline void skb_set_inner_ipproto(struct sk_buff *skb, | |
2854 | __u8 ipproto) | |
2855 | { | |
2856 | skb->inner_ipproto = ipproto; | |
2857 | skb->inner_protocol_type = ENCAP_TYPE_IPPROTO; | |
2858 | } | |
2859 | ||
6a674e9c JG |
2860 | static inline void skb_reset_inner_headers(struct sk_buff *skb) |
2861 | { | |
aefbd2b3 | 2862 | skb->inner_mac_header = skb->mac_header; |
6a674e9c JG |
2863 | skb->inner_network_header = skb->network_header; |
2864 | skb->inner_transport_header = skb->transport_header; | |
2865 | } | |
2866 | ||
0b5c9db1 JP |
2867 | static inline void skb_reset_mac_len(struct sk_buff *skb) |
2868 | { | |
2869 | skb->mac_len = skb->network_header - skb->mac_header; | |
2870 | } | |
2871 | ||
6a674e9c JG |
2872 | static inline unsigned char *skb_inner_transport_header(const struct sk_buff |
2873 | *skb) | |
2874 | { | |
2875 | return skb->head + skb->inner_transport_header; | |
2876 | } | |
2877 | ||
55dc5a9f TH |
2878 | static inline int skb_inner_transport_offset(const struct sk_buff *skb) |
2879 | { | |
2880 | return skb_inner_transport_header(skb) - skb->data; | |
2881 | } | |
2882 | ||
6a674e9c JG |
2883 | static inline void skb_reset_inner_transport_header(struct sk_buff *skb) |
2884 | { | |
2885 | skb->inner_transport_header = skb->data - skb->head; | |
2886 | } | |
2887 | ||
2888 | static inline void skb_set_inner_transport_header(struct sk_buff *skb, | |
2889 | const int offset) | |
2890 | { | |
2891 | skb_reset_inner_transport_header(skb); | |
2892 | skb->inner_transport_header += offset; | |
2893 | } | |
2894 | ||
2895 | static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb) | |
2896 | { | |
2897 | return skb->head + skb->inner_network_header; | |
2898 | } | |
2899 | ||
2900 | static inline void skb_reset_inner_network_header(struct sk_buff *skb) | |
2901 | { | |
2902 | skb->inner_network_header = skb->data - skb->head; | |
2903 | } | |
2904 | ||
2905 | static inline void skb_set_inner_network_header(struct sk_buff *skb, | |
2906 | const int offset) | |
2907 | { | |
2908 | skb_reset_inner_network_header(skb); | |
2909 | skb->inner_network_header += offset; | |
2910 | } | |
2911 | ||
025f8ad2 FW |
2912 | static inline bool skb_inner_network_header_was_set(const struct sk_buff *skb) |
2913 | { | |
2914 | return skb->inner_network_header > 0; | |
2915 | } | |
2916 | ||
aefbd2b3 PS |
2917 | static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb) |
2918 | { | |
2919 | return skb->head + skb->inner_mac_header; | |
2920 | } | |
2921 | ||
2922 | static inline void skb_reset_inner_mac_header(struct sk_buff *skb) | |
2923 | { | |
2924 | skb->inner_mac_header = skb->data - skb->head; | |
2925 | } | |
2926 | ||
2927 | static inline void skb_set_inner_mac_header(struct sk_buff *skb, | |
2928 | const int offset) | |
2929 | { | |
2930 | skb_reset_inner_mac_header(skb); | |
2931 | skb->inner_mac_header += offset; | |
2932 | } | |
fda55eca ED |
2933 | static inline bool skb_transport_header_was_set(const struct sk_buff *skb) |
2934 | { | |
35d04610 | 2935 | return skb->transport_header != (typeof(skb->transport_header))~0U; |
fda55eca ED |
2936 | } |
2937 | ||
9c70220b ACM |
2938 | static inline unsigned char *skb_transport_header(const struct sk_buff *skb) |
2939 | { | |
66e4c8d9 | 2940 | DEBUG_NET_WARN_ON_ONCE(!skb_transport_header_was_set(skb)); |
2e07fa9c | 2941 | return skb->head + skb->transport_header; |
9c70220b ACM |
2942 | } |
2943 | ||
badff6d0 ACM |
2944 | static inline void skb_reset_transport_header(struct sk_buff *skb) |
2945 | { | |
2e07fa9c | 2946 | skb->transport_header = skb->data - skb->head; |
badff6d0 ACM |
2947 | } |
2948 | ||
967b05f6 ACM |
2949 | static inline void skb_set_transport_header(struct sk_buff *skb, |
2950 | const int offset) | |
2951 | { | |
2e07fa9c ACM |
2952 | skb_reset_transport_header(skb); |
2953 | skb->transport_header += offset; | |
ea2ae17d ACM |
2954 | } |
2955 | ||
d56f90a7 ACM |
2956 | static inline unsigned char *skb_network_header(const struct sk_buff *skb) |
2957 | { | |
2e07fa9c | 2958 | return skb->head + skb->network_header; |
d56f90a7 ACM |
2959 | } |
2960 | ||
c1d2bbe1 ACM |
2961 | static inline void skb_reset_network_header(struct sk_buff *skb) |
2962 | { | |
2e07fa9c | 2963 | skb->network_header = skb->data - skb->head; |
c1d2bbe1 ACM |
2964 | } |
2965 | ||
c14d2450 ACM |
2966 | static inline void skb_set_network_header(struct sk_buff *skb, const int offset) |
2967 | { | |
2e07fa9c ACM |
2968 | skb_reset_network_header(skb); |
2969 | skb->network_header += offset; | |
c14d2450 ACM |
2970 | } |
2971 | ||
f9aefd6b ED |
2972 | static inline int skb_mac_header_was_set(const struct sk_buff *skb) |
2973 | { | |
2974 | return skb->mac_header != (typeof(skb->mac_header))~0U; | |
2975 | } | |
2976 | ||
2e07fa9c | 2977 | static inline unsigned char *skb_mac_header(const struct sk_buff *skb) |
bbe735e4 | 2978 | { |
f9aefd6b | 2979 | DEBUG_NET_WARN_ON_ONCE(!skb_mac_header_was_set(skb)); |
2e07fa9c | 2980 | return skb->head + skb->mac_header; |
bbe735e4 ACM |
2981 | } |
2982 | ||
ea6da4fd AV |
2983 | static inline int skb_mac_offset(const struct sk_buff *skb) |
2984 | { | |
2985 | return skb_mac_header(skb) - skb->data; | |
2986 | } | |
2987 | ||
0daf4349 DB |
2988 | static inline u32 skb_mac_header_len(const struct sk_buff *skb) |
2989 | { | |
f9aefd6b | 2990 | DEBUG_NET_WARN_ON_ONCE(!skb_mac_header_was_set(skb)); |
0daf4349 DB |
2991 | return skb->network_header - skb->mac_header; |
2992 | } | |
2993 | ||
b4ab3141 DB |
2994 | static inline void skb_unset_mac_header(struct sk_buff *skb) |
2995 | { | |
2996 | skb->mac_header = (typeof(skb->mac_header))~0U; | |
2997 | } | |
2998 | ||
2e07fa9c ACM |
2999 | static inline void skb_reset_mac_header(struct sk_buff *skb) |
3000 | { | |
3001 | skb->mac_header = skb->data - skb->head; | |
3002 | } | |
3003 | ||
3004 | static inline void skb_set_mac_header(struct sk_buff *skb, const int offset) | |
3005 | { | |
3006 | skb_reset_mac_header(skb); | |
3007 | skb->mac_header += offset; | |
3008 | } | |
3009 | ||
0e3da5bb TT |
3010 | static inline void skb_pop_mac_header(struct sk_buff *skb) |
3011 | { | |
3012 | skb->mac_header = skb->network_header; | |
3013 | } | |
3014 | ||
d2aa125d | 3015 | static inline void skb_probe_transport_header(struct sk_buff *skb) |
fbbdb8f0 | 3016 | { |
72a338bc | 3017 | struct flow_keys_basic keys; |
fbbdb8f0 YX |
3018 | |
3019 | if (skb_transport_header_was_set(skb)) | |
3020 | return; | |
72a338bc | 3021 | |
3cbf4ffb SF |
3022 | if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys, |
3023 | NULL, 0, 0, 0, 0)) | |
42aecaa9 | 3024 | skb_set_transport_header(skb, keys.control.thoff); |
fbbdb8f0 YX |
3025 | } |
3026 | ||
03606895 ED |
3027 | static inline void skb_mac_header_rebuild(struct sk_buff *skb) |
3028 | { | |
3029 | if (skb_mac_header_was_set(skb)) { | |
3030 | const unsigned char *old_mac = skb_mac_header(skb); | |
3031 | ||
3032 | skb_set_mac_header(skb, -skb->mac_len); | |
3033 | memmove(skb_mac_header(skb), old_mac, skb->mac_len); | |
3034 | } | |
3035 | } | |
3036 | ||
58fbfeca PD |
3037 | /* Move the full mac header up to current network_header. |
3038 | * Leaves skb->data pointing at offset skb->mac_len into the mac_header. | |
3039 | * Must be provided the complete mac header length. | |
3040 | */ | |
3041 | static inline void skb_mac_header_rebuild_full(struct sk_buff *skb, u32 full_mac_len) | |
3042 | { | |
3043 | if (skb_mac_header_was_set(skb)) { | |
3044 | const unsigned char *old_mac = skb_mac_header(skb); | |
3045 | ||
3046 | skb_set_mac_header(skb, -full_mac_len); | |
3047 | memmove(skb_mac_header(skb), old_mac, full_mac_len); | |
3048 | __skb_push(skb, full_mac_len - skb->mac_len); | |
3049 | } | |
3050 | } | |
3051 | ||
04fb451e MM |
3052 | static inline int skb_checksum_start_offset(const struct sk_buff *skb) |
3053 | { | |
3054 | return skb->csum_start - skb_headroom(skb); | |
3055 | } | |
3056 | ||
08b64fcc AD |
3057 | static inline unsigned char *skb_checksum_start(const struct sk_buff *skb) |
3058 | { | |
3059 | return skb->head + skb->csum_start; | |
3060 | } | |
3061 | ||
2e07fa9c ACM |
3062 | static inline int skb_transport_offset(const struct sk_buff *skb) |
3063 | { | |
3064 | return skb_transport_header(skb) - skb->data; | |
3065 | } | |
3066 | ||
3067 | static inline u32 skb_network_header_len(const struct sk_buff *skb) | |
3068 | { | |
cc15bd10 | 3069 | DEBUG_NET_WARN_ON_ONCE(!skb_transport_header_was_set(skb)); |
2e07fa9c ACM |
3070 | return skb->transport_header - skb->network_header; |
3071 | } | |
3072 | ||
6a674e9c JG |
3073 | static inline u32 skb_inner_network_header_len(const struct sk_buff *skb) |
3074 | { | |
3075 | return skb->inner_transport_header - skb->inner_network_header; | |
3076 | } | |
3077 | ||
2e07fa9c ACM |
3078 | static inline int skb_network_offset(const struct sk_buff *skb) |
3079 | { | |
3080 | return skb_network_header(skb) - skb->data; | |
3081 | } | |
48d49d0c | 3082 | |
6a674e9c JG |
3083 | static inline int skb_inner_network_offset(const struct sk_buff *skb) |
3084 | { | |
3085 | return skb_inner_network_header(skb) - skb->data; | |
3086 | } | |
3087 | ||
f9599ce1 CG |
3088 | static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len) |
3089 | { | |
3090 | return pskb_may_pull(skb, skb_network_offset(skb) + len); | |
3091 | } | |
3092 | ||
1da177e4 LT |
3093 | /* |
3094 | * CPUs often take a performance hit when accessing unaligned memory | |
3095 | * locations. The actual performance hit varies, it can be small if the | |
3096 | * hardware handles it or large if we have to take an exception and fix it | |
3097 | * in software. | |
3098 | * | |
3099 | * Since an ethernet header is 14 bytes network drivers often end up with | |
3100 | * the IP header at an unaligned offset. The IP header can be aligned by | |
3101 | * shifting the start of the packet by 2 bytes. Drivers should do this | |
3102 | * with: | |
3103 | * | |
8660c124 | 3104 | * skb_reserve(skb, NET_IP_ALIGN); |
1da177e4 LT |
3105 | * |
3106 | * The downside to this alignment of the IP header is that the DMA is now | |
3107 | * unaligned. On some architectures the cost of an unaligned DMA is high | |
3108 | * and this cost outweighs the gains made by aligning the IP header. | |
8660c124 | 3109 | * |
1da177e4 LT |
3110 | * Since this trade off varies between architectures, we allow NET_IP_ALIGN |
3111 | * to be overridden. | |
3112 | */ | |
3113 | #ifndef NET_IP_ALIGN | |
3114 | #define NET_IP_ALIGN 2 | |
3115 | #endif | |
3116 | ||
025be81e AB |
3117 | /* |
3118 | * The networking layer reserves some headroom in skb data (via | |
3119 | * dev_alloc_skb). This is used to avoid having to reallocate skb data when | |
3120 | * the header has to grow. In the default case, if the header has to grow | |
d6301d3d | 3121 | * 32 bytes or less we avoid the reallocation. |
025be81e AB |
3122 | * |
3123 | * Unfortunately this headroom changes the DMA alignment of the resulting | |
3124 | * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive | |
3125 | * on some architectures. An architecture can override this value, | |
3126 | * perhaps setting it to a cacheline in size (since that will maintain | |
3127 | * cacheline alignment of the DMA). It must be a power of 2. | |
3128 | * | |
d6301d3d | 3129 | * Various parts of the networking layer expect at least 32 bytes of |
025be81e | 3130 | * headroom, you should not reduce this. |
5933dd2f ED |
3131 | * |
3132 | * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS) | |
3133 | * to reduce average number of cache lines per packet. | |
645f0897 | 3134 | * get_rps_cpu() for example only access one 64 bytes aligned block : |
18e8c134 | 3135 | * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8) |
025be81e AB |
3136 | */ |
3137 | #ifndef NET_SKB_PAD | |
5933dd2f | 3138 | #define NET_SKB_PAD max(32, L1_CACHE_BYTES) |
025be81e AB |
3139 | #endif |
3140 | ||
7965bd4d | 3141 | int ___pskb_trim(struct sk_buff *skb, unsigned int len); |
1da177e4 | 3142 | |
5293efe6 | 3143 | static inline void __skb_set_length(struct sk_buff *skb, unsigned int len) |
1da177e4 | 3144 | { |
5e1abdc3 | 3145 | if (WARN_ON(skb_is_nonlinear(skb))) |
3cc0e873 | 3146 | return; |
27a884dc ACM |
3147 | skb->len = len; |
3148 | skb_set_tail_pointer(skb, len); | |
1da177e4 LT |
3149 | } |
3150 | ||
5293efe6 DB |
3151 | static inline void __skb_trim(struct sk_buff *skb, unsigned int len) |
3152 | { | |
3153 | __skb_set_length(skb, len); | |
3154 | } | |
3155 | ||
7965bd4d | 3156 | void skb_trim(struct sk_buff *skb, unsigned int len); |
1da177e4 LT |
3157 | |
3158 | static inline int __pskb_trim(struct sk_buff *skb, unsigned int len) | |
3159 | { | |
3cc0e873 HX |
3160 | if (skb->data_len) |
3161 | return ___pskb_trim(skb, len); | |
3162 | __skb_trim(skb, len); | |
3163 | return 0; | |
1da177e4 LT |
3164 | } |
3165 | ||
3166 | static inline int pskb_trim(struct sk_buff *skb, unsigned int len) | |
3167 | { | |
3168 | return (len < skb->len) ? __pskb_trim(skb, len) : 0; | |
3169 | } | |
3170 | ||
e9fa4f7b HX |
3171 | /** |
3172 | * pskb_trim_unique - remove end from a paged unique (not cloned) buffer | |
3173 | * @skb: buffer to alter | |
3174 | * @len: new length | |
3175 | * | |
3176 | * This is identical to pskb_trim except that the caller knows that | |
3177 | * the skb is not cloned so we should never get an error due to out- | |
3178 | * of-memory. | |
3179 | */ | |
3180 | static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len) | |
3181 | { | |
3182 | int err = pskb_trim(skb, len); | |
3183 | BUG_ON(err); | |
3184 | } | |
3185 | ||
5293efe6 DB |
3186 | static inline int __skb_grow(struct sk_buff *skb, unsigned int len) |
3187 | { | |
3188 | unsigned int diff = len - skb->len; | |
3189 | ||
3190 | if (skb_tailroom(skb) < diff) { | |
3191 | int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb), | |
3192 | GFP_ATOMIC); | |
3193 | if (ret) | |
3194 | return ret; | |
3195 | } | |
3196 | __skb_set_length(skb, len); | |
3197 | return 0; | |
3198 | } | |
3199 | ||
1da177e4 LT |
3200 | /** |
3201 | * skb_orphan - orphan a buffer | |
3202 | * @skb: buffer to orphan | |
3203 | * | |
3204 | * If a buffer currently has an owner then we call the owner's | |
3205 | * destructor function and make the @skb unowned. The buffer continues | |
3206 | * to exist but is no longer charged to its former owner. | |
3207 | */ | |
3208 | static inline void skb_orphan(struct sk_buff *skb) | |
3209 | { | |
c34a7612 | 3210 | if (skb->destructor) { |
1da177e4 | 3211 | skb->destructor(skb); |
c34a7612 ED |
3212 | skb->destructor = NULL; |
3213 | skb->sk = NULL; | |
376c7311 ED |
3214 | } else { |
3215 | BUG_ON(skb->sk); | |
c34a7612 | 3216 | } |
1da177e4 LT |
3217 | } |
3218 | ||
a353e0ce MT |
3219 | /** |
3220 | * skb_orphan_frags - orphan the frags contained in a buffer | |
3221 | * @skb: buffer to orphan frags from | |
3222 | * @gfp_mask: allocation mask for replacement pages | |
3223 | * | |
3224 | * For each frag in the SKB which needs a destructor (i.e. has an | |
3225 | * owner) create a copy of that frag and release the original | |
3226 | * page by calling the destructor. | |
3227 | */ | |
3228 | static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask) | |
3229 | { | |
1f8b977a WB |
3230 | if (likely(!skb_zcopy(skb))) |
3231 | return 0; | |
2e07a521 | 3232 | if (skb_shinfo(skb)->flags & SKBFL_DONT_ORPHAN) |
1f8b977a WB |
3233 | return 0; |
3234 | return skb_copy_ubufs(skb, gfp_mask); | |
3235 | } | |
3236 | ||
3237 | /* Frags must be orphaned, even if refcounted, if skb might loop to rx path */ | |
3238 | static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask) | |
3239 | { | |
3240 | if (likely(!skb_zcopy(skb))) | |
a353e0ce MT |
3241 | return 0; |
3242 | return skb_copy_ubufs(skb, gfp_mask); | |
3243 | } | |
3244 | ||
1da177e4 | 3245 | /** |
4025d3e7 | 3246 | * __skb_queue_purge_reason - empty a list |
1da177e4 | 3247 | * @list: list to empty |
4025d3e7 | 3248 | * @reason: drop reason |
1da177e4 LT |
3249 | * |
3250 | * Delete all buffers on an &sk_buff list. Each buffer is removed from | |
3251 | * the list and one reference dropped. This function does not take the | |
3252 | * list lock and the caller must hold the relevant locks to use it. | |
3253 | */ | |
4025d3e7 ED |
3254 | static inline void __skb_queue_purge_reason(struct sk_buff_head *list, |
3255 | enum skb_drop_reason reason) | |
1da177e4 LT |
3256 | { |
3257 | struct sk_buff *skb; | |
4025d3e7 | 3258 | |
1da177e4 | 3259 | while ((skb = __skb_dequeue(list)) != NULL) |
4025d3e7 ED |
3260 | kfree_skb_reason(skb, reason); |
3261 | } | |
3262 | ||
3263 | static inline void __skb_queue_purge(struct sk_buff_head *list) | |
3264 | { | |
3265 | __skb_queue_purge_reason(list, SKB_DROP_REASON_QUEUE_PURGE); | |
3266 | } | |
3267 | ||
3268 | void skb_queue_purge_reason(struct sk_buff_head *list, | |
3269 | enum skb_drop_reason reason); | |
3270 | ||
3271 | static inline void skb_queue_purge(struct sk_buff_head *list) | |
3272 | { | |
3273 | skb_queue_purge_reason(list, SKB_DROP_REASON_QUEUE_PURGE); | |
1da177e4 LT |
3274 | } |
3275 | ||
385114de | 3276 | unsigned int skb_rbtree_purge(struct rb_root *root); |
0f158b32 | 3277 | void skb_errqueue_purge(struct sk_buff_head *list); |
9f5afeae | 3278 | |
3f6e687d KH |
3279 | void *__netdev_alloc_frag_align(unsigned int fragsz, unsigned int align_mask); |
3280 | ||
3281 | /** | |
3282 | * netdev_alloc_frag - allocate a page fragment | |
3283 | * @fragsz: fragment size | |
3284 | * | |
3285 | * Allocates a frag from a page for receive buffer. | |
3286 | * Uses GFP_ATOMIC allocations. | |
3287 | */ | |
3288 | static inline void *netdev_alloc_frag(unsigned int fragsz) | |
3289 | { | |
3290 | return __netdev_alloc_frag_align(fragsz, ~0u); | |
3291 | } | |
3292 | ||
3293 | static inline void *netdev_alloc_frag_align(unsigned int fragsz, | |
3294 | unsigned int align) | |
3295 | { | |
3296 | WARN_ON_ONCE(!is_power_of_2(align)); | |
3297 | return __netdev_alloc_frag_align(fragsz, -align); | |
3298 | } | |
1da177e4 | 3299 | |
7965bd4d JP |
3300 | struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length, |
3301 | gfp_t gfp_mask); | |
8af27456 CH |
3302 | |
3303 | /** | |
3304 | * netdev_alloc_skb - allocate an skbuff for rx on a specific device | |
3305 | * @dev: network device to receive on | |
3306 | * @length: length to allocate | |
3307 | * | |
3308 | * Allocate a new &sk_buff and assign it a usage count of one. The | |
3309 | * buffer has unspecified headroom built in. Users should allocate | |
3310 | * the headroom they think they need without accounting for the | |
3311 | * built in space. The built in space is used for optimisations. | |
3312 | * | |
3313 | * %NULL is returned if there is no free memory. Although this function | |
3314 | * allocates memory it can be called from an interrupt. | |
3315 | */ | |
3316 | static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev, | |
6f532612 | 3317 | unsigned int length) |
8af27456 CH |
3318 | { |
3319 | return __netdev_alloc_skb(dev, length, GFP_ATOMIC); | |
3320 | } | |
3321 | ||
6f532612 ED |
3322 | /* legacy helper around __netdev_alloc_skb() */ |
3323 | static inline struct sk_buff *__dev_alloc_skb(unsigned int length, | |
3324 | gfp_t gfp_mask) | |
3325 | { | |
3326 | return __netdev_alloc_skb(NULL, length, gfp_mask); | |
3327 | } | |
3328 | ||
3329 | /* legacy helper around netdev_alloc_skb() */ | |
3330 | static inline struct sk_buff *dev_alloc_skb(unsigned int length) | |
3331 | { | |
3332 | return netdev_alloc_skb(NULL, length); | |
3333 | } | |
3334 | ||
3335 | ||
4915a0de ED |
3336 | static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev, |
3337 | unsigned int length, gfp_t gfp) | |
61321bbd | 3338 | { |
4915a0de | 3339 | struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp); |
61321bbd ED |
3340 | |
3341 | if (NET_IP_ALIGN && skb) | |
3342 | skb_reserve(skb, NET_IP_ALIGN); | |
3343 | return skb; | |
3344 | } | |
3345 | ||
4915a0de ED |
3346 | static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev, |
3347 | unsigned int length) | |
3348 | { | |
3349 | return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC); | |
3350 | } | |
3351 | ||
181edb2b AD |
3352 | static inline void skb_free_frag(void *addr) |
3353 | { | |
8c2dd3e4 | 3354 | page_frag_free(addr); |
181edb2b AD |
3355 | } |
3356 | ||
3f6e687d KH |
3357 | void *__napi_alloc_frag_align(unsigned int fragsz, unsigned int align_mask); |
3358 | ||
3359 | static inline void *napi_alloc_frag(unsigned int fragsz) | |
3360 | { | |
3361 | return __napi_alloc_frag_align(fragsz, ~0u); | |
3362 | } | |
3363 | ||
3364 | static inline void *napi_alloc_frag_align(unsigned int fragsz, | |
3365 | unsigned int align) | |
3366 | { | |
3367 | WARN_ON_ONCE(!is_power_of_2(align)); | |
3368 | return __napi_alloc_frag_align(fragsz, -align); | |
3369 | } | |
3370 | ||
6e9b0190 | 3371 | struct sk_buff *napi_alloc_skb(struct napi_struct *napi, unsigned int length); |
795bb1c0 JDB |
3372 | void napi_consume_skb(struct sk_buff *skb, int budget); |
3373 | ||
9243adfc | 3374 | void napi_skb_free_stolen_head(struct sk_buff *skb); |
8fa66e4a | 3375 | void __napi_kfree_skb(struct sk_buff *skb, enum skb_drop_reason reason); |
ffde7328 | 3376 | |
71dfda58 AD |
3377 | /** |
3378 | * __dev_alloc_pages - allocate page for network Rx | |
3379 | * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx | |
3380 | * @order: size of the allocation | |
3381 | * | |
3382 | * Allocate a new page. | |
3383 | * | |
3384 | * %NULL is returned if there is no free memory. | |
3385 | */ | |
2c321f3f | 3386 | static inline struct page *__dev_alloc_pages_noprof(gfp_t gfp_mask, |
71dfda58 AD |
3387 | unsigned int order) |
3388 | { | |
3389 | /* This piece of code contains several assumptions. | |
bf873a80 | 3390 | * 1. This is for device Rx, therefore a cold page is preferred. |
71dfda58 AD |
3391 | * 2. The expectation is the user wants a compound page. |
3392 | * 3. If requesting a order 0 page it will not be compound | |
3393 | * due to the check to see if order has a value in prep_new_page | |
3394 | * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to | |
3395 | * code in gfp_to_alloc_flags that should be enforcing this. | |
3396 | */ | |
453f85d4 | 3397 | gfp_mask |= __GFP_COMP | __GFP_MEMALLOC; |
71dfda58 | 3398 | |
2c321f3f | 3399 | return alloc_pages_node_noprof(NUMA_NO_NODE, gfp_mask, order); |
71dfda58 | 3400 | } |
2c321f3f | 3401 | #define __dev_alloc_pages(...) alloc_hooks(__dev_alloc_pages_noprof(__VA_ARGS__)) |
71dfda58 | 3402 | |
2c321f3f | 3403 | #define dev_alloc_pages(_order) __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, _order) |
71dfda58 AD |
3404 | |
3405 | /** | |
3406 | * __dev_alloc_page - allocate a page for network Rx | |
3407 | * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx | |
3408 | * | |
3409 | * Allocate a new page. | |
3410 | * | |
3411 | * %NULL is returned if there is no free memory. | |
3412 | */ | |
2c321f3f | 3413 | static inline struct page *__dev_alloc_page_noprof(gfp_t gfp_mask) |
71dfda58 | 3414 | { |
2c321f3f | 3415 | return __dev_alloc_pages_noprof(gfp_mask, 0); |
71dfda58 | 3416 | } |
2c321f3f | 3417 | #define __dev_alloc_page(...) alloc_hooks(__dev_alloc_page_noprof(__VA_ARGS__)) |
71dfda58 | 3418 | |
2c321f3f | 3419 | #define dev_alloc_page() dev_alloc_pages(0) |
71dfda58 | 3420 | |
bc38f30f AL |
3421 | /** |
3422 | * dev_page_is_reusable - check whether a page can be reused for network Rx | |
3423 | * @page: the page to test | |
3424 | * | |
3425 | * A page shouldn't be considered for reusing/recycling if it was allocated | |
3426 | * under memory pressure or at a distant memory node. | |
3427 | * | |
3428 | * Returns false if this page should be returned to page allocator, true | |
3429 | * otherwise. | |
3430 | */ | |
3431 | static inline bool dev_page_is_reusable(const struct page *page) | |
3432 | { | |
3433 | return likely(page_to_nid(page) == numa_mem_id() && | |
3434 | !page_is_pfmemalloc(page)); | |
3435 | } | |
3436 | ||
0614002b MG |
3437 | /** |
3438 | * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page | |
3439 | * @page: The page that was allocated from skb_alloc_page | |
3440 | * @skb: The skb that may need pfmemalloc set | |
3441 | */ | |
48f971c9 AL |
3442 | static inline void skb_propagate_pfmemalloc(const struct page *page, |
3443 | struct sk_buff *skb) | |
0614002b | 3444 | { |
2f064f34 | 3445 | if (page_is_pfmemalloc(page)) |
0614002b MG |
3446 | skb->pfmemalloc = true; |
3447 | } | |
3448 | ||
7240b60c JL |
3449 | /** |
3450 | * skb_frag_off() - Returns the offset of a skb fragment | |
3451 | * @frag: the paged fragment | |
3452 | */ | |
3453 | static inline unsigned int skb_frag_off(const skb_frag_t *frag) | |
3454 | { | |
21d2e673 | 3455 | return frag->offset; |
7240b60c JL |
3456 | } |
3457 | ||
3458 | /** | |
3459 | * skb_frag_off_add() - Increments the offset of a skb fragment by @delta | |
3460 | * @frag: skb fragment | |
3461 | * @delta: value to add | |
3462 | */ | |
3463 | static inline void skb_frag_off_add(skb_frag_t *frag, int delta) | |
3464 | { | |
21d2e673 | 3465 | frag->offset += delta; |
7240b60c JL |
3466 | } |
3467 | ||
3468 | /** | |
3469 | * skb_frag_off_set() - Sets the offset of a skb fragment | |
3470 | * @frag: skb fragment | |
3471 | * @offset: offset of fragment | |
3472 | */ | |
3473 | static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset) | |
3474 | { | |
21d2e673 | 3475 | frag->offset = offset; |
7240b60c JL |
3476 | } |
3477 | ||
3478 | /** | |
3479 | * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment | |
3480 | * @fragto: skb fragment where offset is set | |
3481 | * @fragfrom: skb fragment offset is copied from | |
3482 | */ | |
3483 | static inline void skb_frag_off_copy(skb_frag_t *fragto, | |
3484 | const skb_frag_t *fragfrom) | |
3485 | { | |
21d2e673 | 3486 | fragto->offset = fragfrom->offset; |
7240b60c JL |
3487 | } |
3488 | ||
131ea667 | 3489 | /** |
e227867f | 3490 | * skb_frag_page - retrieve the page referred to by a paged fragment |
131ea667 IC |
3491 | * @frag: the paged fragment |
3492 | * | |
3493 | * Returns the &struct page associated with @frag. | |
3494 | */ | |
3495 | static inline struct page *skb_frag_page(const skb_frag_t *frag) | |
3496 | { | |
21d2e673 | 3497 | return netmem_to_page(frag->netmem); |
131ea667 IC |
3498 | } |
3499 | ||
27accb3c LB |
3500 | int skb_pp_cow_data(struct page_pool *pool, struct sk_buff **pskb, |
3501 | unsigned int headroom); | |
e6d5dbdd LB |
3502 | int skb_cow_data_for_xdp(struct page_pool *pool, struct sk_buff **pskb, |
3503 | struct bpf_prog *prog); | |
131ea667 IC |
3504 | /** |
3505 | * skb_frag_address - gets the address of the data contained in a paged fragment | |
3506 | * @frag: the paged fragment buffer | |
3507 | * | |
3508 | * Returns the address of the data within @frag. The page must already | |
3509 | * be mapped. | |
3510 | */ | |
3511 | static inline void *skb_frag_address(const skb_frag_t *frag) | |
3512 | { | |
7240b60c | 3513 | return page_address(skb_frag_page(frag)) + skb_frag_off(frag); |
131ea667 IC |
3514 | } |
3515 | ||
3516 | /** | |
3517 | * skb_frag_address_safe - gets the address of the data contained in a paged fragment | |
3518 | * @frag: the paged fragment buffer | |
3519 | * | |
3520 | * Returns the address of the data within @frag. Checks that the page | |
3521 | * is mapped and returns %NULL otherwise. | |
3522 | */ | |
3523 | static inline void *skb_frag_address_safe(const skb_frag_t *frag) | |
3524 | { | |
3525 | void *ptr = page_address(skb_frag_page(frag)); | |
3526 | if (unlikely(!ptr)) | |
3527 | return NULL; | |
3528 | ||
7240b60c JL |
3529 | return ptr + skb_frag_off(frag); |
3530 | } | |
3531 | ||
3532 | /** | |
3533 | * skb_frag_page_copy() - sets the page in a fragment from another fragment | |
3534 | * @fragto: skb fragment where page is set | |
3535 | * @fragfrom: skb fragment page is copied from | |
3536 | */ | |
3537 | static inline void skb_frag_page_copy(skb_frag_t *fragto, | |
3538 | const skb_frag_t *fragfrom) | |
3539 | { | |
21d2e673 | 3540 | fragto->netmem = fragfrom->netmem; |
131ea667 IC |
3541 | } |
3542 | ||
400dfd3a ED |
3543 | bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio); |
3544 | ||
131ea667 IC |
3545 | /** |
3546 | * skb_frag_dma_map - maps a paged fragment via the DMA API | |
f83347df | 3547 | * @dev: the device to map the fragment to |
131ea667 IC |
3548 | * @frag: the paged fragment to map |
3549 | * @offset: the offset within the fragment (starting at the | |
3550 | * fragment's own offset) | |
3551 | * @size: the number of bytes to map | |
771b00a8 | 3552 | * @dir: the direction of the mapping (``PCI_DMA_*``) |
131ea667 IC |
3553 | * |
3554 | * Maps the page associated with @frag to @device. | |
3555 | */ | |
3556 | static inline dma_addr_t skb_frag_dma_map(struct device *dev, | |
3557 | const skb_frag_t *frag, | |
3558 | size_t offset, size_t size, | |
3559 | enum dma_data_direction dir) | |
3560 | { | |
3561 | return dma_map_page(dev, skb_frag_page(frag), | |
7240b60c | 3562 | skb_frag_off(frag) + offset, size, dir); |
131ea667 IC |
3563 | } |
3564 | ||
117632e6 ED |
3565 | static inline struct sk_buff *pskb_copy(struct sk_buff *skb, |
3566 | gfp_t gfp_mask) | |
3567 | { | |
3568 | return __pskb_copy(skb, skb_headroom(skb), gfp_mask); | |
3569 | } | |
3570 | ||
bad93e9d OP |
3571 | |
3572 | static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb, | |
3573 | gfp_t gfp_mask) | |
3574 | { | |
3575 | return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true); | |
3576 | } | |
3577 | ||
3578 | ||
334a8132 PM |
3579 | /** |
3580 | * skb_clone_writable - is the header of a clone writable | |
3581 | * @skb: buffer to check | |
3582 | * @len: length up to which to write | |
3583 | * | |
3584 | * Returns true if modifying the header part of the cloned buffer | |
3585 | * does not requires the data to be copied. | |
3586 | */ | |
05bdd2f1 | 3587 | static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len) |
334a8132 PM |
3588 | { |
3589 | return !skb_header_cloned(skb) && | |
3590 | skb_headroom(skb) + len <= skb->hdr_len; | |
3591 | } | |
3592 | ||
3697649f DB |
3593 | static inline int skb_try_make_writable(struct sk_buff *skb, |
3594 | unsigned int write_len) | |
3595 | { | |
3596 | return skb_cloned(skb) && !skb_clone_writable(skb, write_len) && | |
3597 | pskb_expand_head(skb, 0, 0, GFP_ATOMIC); | |
3598 | } | |
3599 | ||
d9cc2048 HX |
3600 | static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom, |
3601 | int cloned) | |
3602 | { | |
3603 | int delta = 0; | |
3604 | ||
d9cc2048 HX |
3605 | if (headroom > skb_headroom(skb)) |
3606 | delta = headroom - skb_headroom(skb); | |
3607 | ||
3608 | if (delta || cloned) | |
3609 | return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0, | |
3610 | GFP_ATOMIC); | |
3611 | return 0; | |
3612 | } | |
3613 | ||
1da177e4 LT |
3614 | /** |
3615 | * skb_cow - copy header of skb when it is required | |
3616 | * @skb: buffer to cow | |
3617 | * @headroom: needed headroom | |
3618 | * | |
3619 | * If the skb passed lacks sufficient headroom or its data part | |
3620 | * is shared, data is reallocated. If reallocation fails, an error | |
3621 | * is returned and original skb is not changed. | |
3622 | * | |
3623 | * The result is skb with writable area skb->head...skb->tail | |
3624 | * and at least @headroom of space at head. | |
3625 | */ | |
3626 | static inline int skb_cow(struct sk_buff *skb, unsigned int headroom) | |
3627 | { | |
d9cc2048 HX |
3628 | return __skb_cow(skb, headroom, skb_cloned(skb)); |
3629 | } | |
1da177e4 | 3630 | |
d9cc2048 HX |
3631 | /** |
3632 | * skb_cow_head - skb_cow but only making the head writable | |
3633 | * @skb: buffer to cow | |
3634 | * @headroom: needed headroom | |
3635 | * | |
3636 | * This function is identical to skb_cow except that we replace the | |
3637 | * skb_cloned check by skb_header_cloned. It should be used when | |
3638 | * you only need to push on some header and do not need to modify | |
3639 | * the data. | |
3640 | */ | |
3641 | static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom) | |
3642 | { | |
3643 | return __skb_cow(skb, headroom, skb_header_cloned(skb)); | |
1da177e4 LT |
3644 | } |
3645 | ||
3646 | /** | |
3647 | * skb_padto - pad an skbuff up to a minimal size | |
3648 | * @skb: buffer to pad | |
3649 | * @len: minimal length | |
3650 | * | |
3651 | * Pads up a buffer to ensure the trailing bytes exist and are | |
3652 | * blanked. If the buffer already contains sufficient data it | |
5b057c6b HX |
3653 | * is untouched. Otherwise it is extended. Returns zero on |
3654 | * success. The skb is freed on error. | |
1da177e4 | 3655 | */ |
5b057c6b | 3656 | static inline int skb_padto(struct sk_buff *skb, unsigned int len) |
1da177e4 LT |
3657 | { |
3658 | unsigned int size = skb->len; | |
3659 | if (likely(size >= len)) | |
5b057c6b | 3660 | return 0; |
987c402a | 3661 | return skb_pad(skb, len - size); |
1da177e4 LT |
3662 | } |
3663 | ||
9c0c1124 | 3664 | /** |
4ea7b0cf | 3665 | * __skb_put_padto - increase size and pad an skbuff up to a minimal size |
9c0c1124 AD |
3666 | * @skb: buffer to pad |
3667 | * @len: minimal length | |
cd0a137a | 3668 | * @free_on_error: free buffer on error |
9c0c1124 AD |
3669 | * |
3670 | * Pads up a buffer to ensure the trailing bytes exist and are | |
3671 | * blanked. If the buffer already contains sufficient data it | |
3672 | * is untouched. Otherwise it is extended. Returns zero on | |
cd0a137a | 3673 | * success. The skb is freed on error if @free_on_error is true. |
9c0c1124 | 3674 | */ |
4a009cb0 ED |
3675 | static inline int __must_check __skb_put_padto(struct sk_buff *skb, |
3676 | unsigned int len, | |
3677 | bool free_on_error) | |
9c0c1124 AD |
3678 | { |
3679 | unsigned int size = skb->len; | |
3680 | ||
3681 | if (unlikely(size < len)) { | |
3682 | len -= size; | |
cd0a137a | 3683 | if (__skb_pad(skb, len, free_on_error)) |
9c0c1124 AD |
3684 | return -ENOMEM; |
3685 | __skb_put(skb, len); | |
3686 | } | |
3687 | return 0; | |
3688 | } | |
3689 | ||
cd0a137a FF |
3690 | /** |
3691 | * skb_put_padto - increase size and pad an skbuff up to a minimal size | |
3692 | * @skb: buffer to pad | |
3693 | * @len: minimal length | |
3694 | * | |
3695 | * Pads up a buffer to ensure the trailing bytes exist and are | |
3696 | * blanked. If the buffer already contains sufficient data it | |
3697 | * is untouched. Otherwise it is extended. Returns zero on | |
3698 | * success. The skb is freed on error. | |
3699 | */ | |
4a009cb0 | 3700 | static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len) |
cd0a137a FF |
3701 | { |
3702 | return __skb_put_padto(skb, len, true); | |
3703 | } | |
3704 | ||
7c6f353e DH |
3705 | bool csum_and_copy_from_iter_full(void *addr, size_t bytes, __wsum *csum, struct iov_iter *i) |
3706 | __must_check; | |
6d0d4199 | 3707 | |
1da177e4 | 3708 | static inline int skb_add_data(struct sk_buff *skb, |
af2b040e | 3709 | struct iov_iter *from, int copy) |
1da177e4 LT |
3710 | { |
3711 | const int off = skb->len; | |
3712 | ||
3713 | if (skb->ip_summed == CHECKSUM_NONE) { | |
af2b040e | 3714 | __wsum csum = 0; |
15e6cb46 AV |
3715 | if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy, |
3716 | &csum, from)) { | |
1da177e4 LT |
3717 | skb->csum = csum_block_add(skb->csum, csum, off); |
3718 | return 0; | |
3719 | } | |
15e6cb46 | 3720 | } else if (copy_from_iter_full(skb_put(skb, copy), copy, from)) |
1da177e4 LT |
3721 | return 0; |
3722 | ||
3723 | __skb_trim(skb, off); | |
3724 | return -EFAULT; | |
3725 | } | |
3726 | ||
38ba0a65 ED |
3727 | static inline bool skb_can_coalesce(struct sk_buff *skb, int i, |
3728 | const struct page *page, int off) | |
1da177e4 | 3729 | { |
1f8b977a WB |
3730 | if (skb_zcopy(skb)) |
3731 | return false; | |
1da177e4 | 3732 | if (i) { |
d8e18a51 | 3733 | const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1]; |
1da177e4 | 3734 | |
ea2ab693 | 3735 | return page == skb_frag_page(frag) && |
7240b60c | 3736 | off == skb_frag_off(frag) + skb_frag_size(frag); |
1da177e4 | 3737 | } |
38ba0a65 | 3738 | return false; |
1da177e4 LT |
3739 | } |
3740 | ||
364c6bad HX |
3741 | static inline int __skb_linearize(struct sk_buff *skb) |
3742 | { | |
3743 | return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM; | |
3744 | } | |
3745 | ||
1da177e4 LT |
3746 | /** |
3747 | * skb_linearize - convert paged skb to linear one | |
3748 | * @skb: buffer to linarize | |
1da177e4 LT |
3749 | * |
3750 | * If there is no free memory -ENOMEM is returned, otherwise zero | |
3751 | * is returned and the old skb data released. | |
3752 | */ | |
364c6bad HX |
3753 | static inline int skb_linearize(struct sk_buff *skb) |
3754 | { | |
3755 | return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0; | |
3756 | } | |
3757 | ||
cef401de ED |
3758 | /** |
3759 | * skb_has_shared_frag - can any frag be overwritten | |
3760 | * @skb: buffer to test | |
3761 | * | |
3762 | * Return true if the skb has at least one frag that might be modified | |
3763 | * by an external entity (as in vmsplice()/sendfile()) | |
3764 | */ | |
3765 | static inline bool skb_has_shared_frag(const struct sk_buff *skb) | |
3766 | { | |
c9af6db4 | 3767 | return skb_is_nonlinear(skb) && |
06b4feb3 | 3768 | skb_shinfo(skb)->flags & SKBFL_SHARED_FRAG; |
cef401de ED |
3769 | } |
3770 | ||
364c6bad HX |
3771 | /** |
3772 | * skb_linearize_cow - make sure skb is linear and writable | |
3773 | * @skb: buffer to process | |
3774 | * | |
3775 | * If there is no free memory -ENOMEM is returned, otherwise zero | |
3776 | * is returned and the old skb data released. | |
3777 | */ | |
3778 | static inline int skb_linearize_cow(struct sk_buff *skb) | |
1da177e4 | 3779 | { |
364c6bad HX |
3780 | return skb_is_nonlinear(skb) || skb_cloned(skb) ? |
3781 | __skb_linearize(skb) : 0; | |
1da177e4 LT |
3782 | } |
3783 | ||
479ffccc DB |
3784 | static __always_inline void |
3785 | __skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len, | |
3786 | unsigned int off) | |
3787 | { | |
3788 | if (skb->ip_summed == CHECKSUM_COMPLETE) | |
3789 | skb->csum = csum_block_sub(skb->csum, | |
3790 | csum_partial(start, len, 0), off); | |
3791 | else if (skb->ip_summed == CHECKSUM_PARTIAL && | |
3792 | skb_checksum_start_offset(skb) < 0) | |
3793 | skb->ip_summed = CHECKSUM_NONE; | |
3794 | } | |
3795 | ||
1da177e4 LT |
3796 | /** |
3797 | * skb_postpull_rcsum - update checksum for received skb after pull | |
3798 | * @skb: buffer to update | |
3799 | * @start: start of data before pull | |
3800 | * @len: length of data pulled | |
3801 | * | |
3802 | * After doing a pull on a received packet, you need to call this to | |
84fa7933 PM |
3803 | * update the CHECKSUM_COMPLETE checksum, or set ip_summed to |
3804 | * CHECKSUM_NONE so that it can be recomputed from scratch. | |
1da177e4 | 3805 | */ |
1da177e4 | 3806 | static inline void skb_postpull_rcsum(struct sk_buff *skb, |
cbb042f9 | 3807 | const void *start, unsigned int len) |
1da177e4 | 3808 | { |
29c30026 | 3809 | if (skb->ip_summed == CHECKSUM_COMPLETE) |
45cac675 ED |
3810 | skb->csum = wsum_negate(csum_partial(start, len, |
3811 | wsum_negate(skb->csum))); | |
29c30026 ED |
3812 | else if (skb->ip_summed == CHECKSUM_PARTIAL && |
3813 | skb_checksum_start_offset(skb) < 0) | |
3814 | skb->ip_summed = CHECKSUM_NONE; | |
1da177e4 LT |
3815 | } |
3816 | ||
479ffccc DB |
3817 | static __always_inline void |
3818 | __skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len, | |
3819 | unsigned int off) | |
3820 | { | |
3821 | if (skb->ip_summed == CHECKSUM_COMPLETE) | |
3822 | skb->csum = csum_block_add(skb->csum, | |
3823 | csum_partial(start, len, 0), off); | |
3824 | } | |
cbb042f9 | 3825 | |
479ffccc DB |
3826 | /** |
3827 | * skb_postpush_rcsum - update checksum for received skb after push | |
3828 | * @skb: buffer to update | |
3829 | * @start: start of data after push | |
3830 | * @len: length of data pushed | |
3831 | * | |
3832 | * After doing a push on a received packet, you need to call this to | |
3833 | * update the CHECKSUM_COMPLETE checksum. | |
3834 | */ | |
f8ffad69 DB |
3835 | static inline void skb_postpush_rcsum(struct sk_buff *skb, |
3836 | const void *start, unsigned int len) | |
3837 | { | |
479ffccc | 3838 | __skb_postpush_rcsum(skb, start, len, 0); |
f8ffad69 DB |
3839 | } |
3840 | ||
af72868b | 3841 | void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len); |
479ffccc | 3842 | |
82a31b92 WC |
3843 | /** |
3844 | * skb_push_rcsum - push skb and update receive checksum | |
3845 | * @skb: buffer to update | |
3846 | * @len: length of data pulled | |
3847 | * | |
3848 | * This function performs an skb_push on the packet and updates | |
3849 | * the CHECKSUM_COMPLETE checksum. It should be used on | |
3850 | * receive path processing instead of skb_push unless you know | |
3851 | * that the checksum difference is zero (e.g., a valid IP header) | |
3852 | * or you are setting ip_summed to CHECKSUM_NONE. | |
3853 | */ | |
d58ff351 | 3854 | static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len) |
82a31b92 WC |
3855 | { |
3856 | skb_push(skb, len); | |
3857 | skb_postpush_rcsum(skb, skb->data, len); | |
3858 | return skb->data; | |
3859 | } | |
3860 | ||
88078d98 | 3861 | int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len); |
7ce5a27f DM |
3862 | /** |
3863 | * pskb_trim_rcsum - trim received skb and update checksum | |
3864 | * @skb: buffer to trim | |
3865 | * @len: new length | |
3866 | * | |
3867 | * This is exactly the same as pskb_trim except that it ensures the | |
3868 | * checksum of received packets are still valid after the operation. | |
6c57f045 | 3869 | * It can change skb pointers. |
7ce5a27f DM |
3870 | */ |
3871 | ||
3872 | static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len) | |
3873 | { | |
3874 | if (likely(len >= skb->len)) | |
3875 | return 0; | |
88078d98 | 3876 | return pskb_trim_rcsum_slow(skb, len); |
7ce5a27f DM |
3877 | } |
3878 | ||
5293efe6 DB |
3879 | static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len) |
3880 | { | |
3881 | if (skb->ip_summed == CHECKSUM_COMPLETE) | |
3882 | skb->ip_summed = CHECKSUM_NONE; | |
3883 | __skb_trim(skb, len); | |
3884 | return 0; | |
3885 | } | |
3886 | ||
3887 | static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len) | |
3888 | { | |
3889 | if (skb->ip_summed == CHECKSUM_COMPLETE) | |
3890 | skb->ip_summed = CHECKSUM_NONE; | |
3891 | return __skb_grow(skb, len); | |
3892 | } | |
3893 | ||
18a4c0ea ED |
3894 | #define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode) |
3895 | #define skb_rb_first(root) rb_to_skb(rb_first(root)) | |
3896 | #define skb_rb_last(root) rb_to_skb(rb_last(root)) | |
3897 | #define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode)) | |
3898 | #define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode)) | |
3899 | ||
1da177e4 LT |
3900 | #define skb_queue_walk(queue, skb) \ |
3901 | for (skb = (queue)->next; \ | |
a1e4891f | 3902 | skb != (struct sk_buff *)(queue); \ |
1da177e4 LT |
3903 | skb = skb->next) |
3904 | ||
46f8914e JC |
3905 | #define skb_queue_walk_safe(queue, skb, tmp) \ |
3906 | for (skb = (queue)->next, tmp = skb->next; \ | |
3907 | skb != (struct sk_buff *)(queue); \ | |
3908 | skb = tmp, tmp = skb->next) | |
3909 | ||
1164f52a | 3910 | #define skb_queue_walk_from(queue, skb) \ |
a1e4891f | 3911 | for (; skb != (struct sk_buff *)(queue); \ |
1164f52a DM |
3912 | skb = skb->next) |
3913 | ||
18a4c0ea ED |
3914 | #define skb_rbtree_walk(skb, root) \ |
3915 | for (skb = skb_rb_first(root); skb != NULL; \ | |
3916 | skb = skb_rb_next(skb)) | |
3917 | ||
3918 | #define skb_rbtree_walk_from(skb) \ | |
3919 | for (; skb != NULL; \ | |
3920 | skb = skb_rb_next(skb)) | |
3921 | ||
3922 | #define skb_rbtree_walk_from_safe(skb, tmp) \ | |
3923 | for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \ | |
3924 | skb = tmp) | |
3925 | ||
1164f52a DM |
3926 | #define skb_queue_walk_from_safe(queue, skb, tmp) \ |
3927 | for (tmp = skb->next; \ | |
3928 | skb != (struct sk_buff *)(queue); \ | |
3929 | skb = tmp, tmp = skb->next) | |
3930 | ||
300ce174 SH |
3931 | #define skb_queue_reverse_walk(queue, skb) \ |
3932 | for (skb = (queue)->prev; \ | |
a1e4891f | 3933 | skb != (struct sk_buff *)(queue); \ |
300ce174 SH |
3934 | skb = skb->prev) |
3935 | ||
686a2955 DM |
3936 | #define skb_queue_reverse_walk_safe(queue, skb, tmp) \ |
3937 | for (skb = (queue)->prev, tmp = skb->prev; \ | |
3938 | skb != (struct sk_buff *)(queue); \ | |
3939 | skb = tmp, tmp = skb->prev) | |
3940 | ||
3941 | #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \ | |
3942 | for (tmp = skb->prev; \ | |
3943 | skb != (struct sk_buff *)(queue); \ | |
3944 | skb = tmp, tmp = skb->prev) | |
1da177e4 | 3945 | |
21dc3301 | 3946 | static inline bool skb_has_frag_list(const struct sk_buff *skb) |
ee039871 DM |
3947 | { |
3948 | return skb_shinfo(skb)->frag_list != NULL; | |
3949 | } | |
3950 | ||
3951 | static inline void skb_frag_list_init(struct sk_buff *skb) | |
3952 | { | |
3953 | skb_shinfo(skb)->frag_list = NULL; | |
3954 | } | |
3955 | ||
ee039871 DM |
3956 | #define skb_walk_frags(skb, iter) \ |
3957 | for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next) | |
3958 | ||
ea3793ee | 3959 | |
b50b0580 SD |
3960 | int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue, |
3961 | int *err, long *timeo_p, | |
ea3793ee | 3962 | const struct sk_buff *skb); |
65101aec PA |
3963 | struct sk_buff *__skb_try_recv_from_queue(struct sock *sk, |
3964 | struct sk_buff_head *queue, | |
3965 | unsigned int flags, | |
fd69c399 | 3966 | int *off, int *err, |
65101aec | 3967 | struct sk_buff **last); |
b50b0580 SD |
3968 | struct sk_buff *__skb_try_recv_datagram(struct sock *sk, |
3969 | struct sk_buff_head *queue, | |
e427cad6 | 3970 | unsigned int flags, int *off, int *err, |
ea3793ee | 3971 | struct sk_buff **last); |
b50b0580 SD |
3972 | struct sk_buff *__skb_recv_datagram(struct sock *sk, |
3973 | struct sk_buff_head *sk_queue, | |
e427cad6 | 3974 | unsigned int flags, int *off, int *err); |
f4b41f06 | 3975 | struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned int flags, int *err); |
a11e1d43 LT |
3976 | __poll_t datagram_poll(struct file *file, struct socket *sock, |
3977 | struct poll_table_struct *wait); | |
c0371da6 AV |
3978 | int skb_copy_datagram_iter(const struct sk_buff *from, int offset, |
3979 | struct iov_iter *to, int size); | |
51f3d02b DM |
3980 | static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset, |
3981 | struct msghdr *msg, int size) | |
3982 | { | |
e5a4b0bb | 3983 | return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size); |
227158db | 3984 | } |
e5a4b0bb AV |
3985 | int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen, |
3986 | struct msghdr *msg); | |
65d69e25 SG |
3987 | int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset, |
3988 | struct iov_iter *to, int len, | |
3989 | struct ahash_request *hash); | |
3a654f97 AV |
3990 | int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset, |
3991 | struct iov_iter *from, int len); | |
3a654f97 | 3992 | int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm); |
7965bd4d | 3993 | void skb_free_datagram(struct sock *sk, struct sk_buff *skb); |
7965bd4d | 3994 | int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags); |
7965bd4d JP |
3995 | int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len); |
3996 | int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len); | |
3997 | __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to, | |
8d5930df | 3998 | int len); |
a60e3cc7 | 3999 | int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset, |
7965bd4d | 4000 | struct pipe_inode_info *pipe, unsigned int len, |
25869262 | 4001 | unsigned int flags); |
20bf50de TH |
4002 | int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset, |
4003 | int len); | |
0739cd28 | 4004 | int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len); |
7965bd4d | 4005 | void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to); |
af2806f8 | 4006 | unsigned int skb_zerocopy_headlen(const struct sk_buff *from); |
36d5fe6a ZK |
4007 | int skb_zerocopy(struct sk_buff *to, struct sk_buff *from, |
4008 | int len, int hlen); | |
7965bd4d JP |
4009 | void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len); |
4010 | int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen); | |
4011 | void skb_scrub_packet(struct sk_buff *skb, bool xnet); | |
7965bd4d | 4012 | struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features); |
3a1296a3 SK |
4013 | struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features, |
4014 | unsigned int offset); | |
0d5501c1 | 4015 | struct sk_buff *skb_vlan_untag(struct sk_buff *skb); |
92ece280 | 4016 | int skb_ensure_writable(struct sk_buff *skb, unsigned int write_len); |
90abde49 | 4017 | int skb_ensure_writable_head_tail(struct sk_buff *skb, struct net_device *dev); |
bfca4c52 | 4018 | int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci); |
93515d53 JP |
4019 | int skb_vlan_pop(struct sk_buff *skb); |
4020 | int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci); | |
19fbcb36 GN |
4021 | int skb_eth_pop(struct sk_buff *skb); |
4022 | int skb_eth_push(struct sk_buff *skb, const unsigned char *dst, | |
4023 | const unsigned char *src); | |
fa4e0f88 | 4024 | int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto, |
d04ac224 | 4025 | int mac_len, bool ethernet); |
040b5cfb MV |
4026 | int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len, |
4027 | bool ethernet); | |
d27cf5c5 | 4028 | int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse); |
2a2ea508 | 4029 | int skb_mpls_dec_ttl(struct sk_buff *skb); |
6fa01ccd SV |
4030 | struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy, |
4031 | gfp_t gfp); | |
20380731 | 4032 | |
6ce8e9ce AV |
4033 | static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len) |
4034 | { | |
3073f070 | 4035 | return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT; |
6ce8e9ce AV |
4036 | } |
4037 | ||
7eab8d9e AV |
4038 | static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len) |
4039 | { | |
e5a4b0bb | 4040 | return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT; |
7eab8d9e AV |
4041 | } |
4042 | ||
2817a336 DB |
4043 | struct skb_checksum_ops { |
4044 | __wsum (*update)(const void *mem, int len, __wsum wsum); | |
4045 | __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len); | |
4046 | }; | |
4047 | ||
9617813d DC |
4048 | extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly; |
4049 | ||
2817a336 DB |
4050 | __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len, |
4051 | __wsum csum, const struct skb_checksum_ops *ops); | |
4052 | __wsum skb_checksum(const struct sk_buff *skb, int offset, int len, | |
4053 | __wsum csum); | |
4054 | ||
1e98a0f0 | 4055 | static inline void * __must_check |
e3305138 AL |
4056 | __skb_header_pointer(const struct sk_buff *skb, int offset, int len, |
4057 | const void *data, int hlen, void *buffer) | |
1da177e4 | 4058 | { |
d206121f | 4059 | if (likely(hlen - offset >= len)) |
e3305138 | 4060 | return (void *)data + offset; |
1da177e4 | 4061 | |
6f5a630d | 4062 | if (!skb || unlikely(skb_copy_bits(skb, offset, buffer, len) < 0)) |
1da177e4 LT |
4063 | return NULL; |
4064 | ||
4065 | return buffer; | |
4066 | } | |
4067 | ||
1e98a0f0 ED |
4068 | static inline void * __must_check |
4069 | skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer) | |
690e36e7 DM |
4070 | { |
4071 | return __skb_header_pointer(skb, offset, len, skb->data, | |
4072 | skb_headlen(skb), buffer); | |
4073 | } | |
4074 | ||
6f5a630d AS |
4075 | static inline void * __must_check |
4076 | skb_pointer_if_linear(const struct sk_buff *skb, int offset, int len) | |
4077 | { | |
4078 | if (likely(skb_headlen(skb) - offset >= len)) | |
4079 | return skb->data + offset; | |
4080 | return NULL; | |
4081 | } | |
4082 | ||
4262e5cc DB |
4083 | /** |
4084 | * skb_needs_linearize - check if we need to linearize a given skb | |
4085 | * depending on the given device features. | |
4086 | * @skb: socket buffer to check | |
4087 | * @features: net device features | |
4088 | * | |
4089 | * Returns true if either: | |
4090 | * 1. skb has frag_list and the device doesn't support FRAGLIST, or | |
4091 | * 2. skb is fragmented and the device does not support SG. | |
4092 | */ | |
4093 | static inline bool skb_needs_linearize(struct sk_buff *skb, | |
4094 | netdev_features_t features) | |
4095 | { | |
4096 | return skb_is_nonlinear(skb) && | |
4097 | ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) || | |
4098 | (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG))); | |
4099 | } | |
4100 | ||
d626f62b ACM |
4101 | static inline void skb_copy_from_linear_data(const struct sk_buff *skb, |
4102 | void *to, | |
4103 | const unsigned int len) | |
4104 | { | |
4105 | memcpy(to, skb->data, len); | |
4106 | } | |
4107 | ||
4108 | static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb, | |
4109 | const int offset, void *to, | |
4110 | const unsigned int len) | |
4111 | { | |
4112 | memcpy(to, skb->data + offset, len); | |
4113 | } | |
4114 | ||
27d7ff46 ACM |
4115 | static inline void skb_copy_to_linear_data(struct sk_buff *skb, |
4116 | const void *from, | |
4117 | const unsigned int len) | |
4118 | { | |
4119 | memcpy(skb->data, from, len); | |
4120 | } | |
4121 | ||
4122 | static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb, | |
4123 | const int offset, | |
4124 | const void *from, | |
4125 | const unsigned int len) | |
4126 | { | |
4127 | memcpy(skb->data + offset, from, len); | |
4128 | } | |
4129 | ||
7965bd4d | 4130 | void skb_init(void); |
1da177e4 | 4131 | |
ac45f602 PO |
4132 | static inline ktime_t skb_get_ktime(const struct sk_buff *skb) |
4133 | { | |
4134 | return skb->tstamp; | |
4135 | } | |
4136 | ||
a61bbcf2 PM |
4137 | /** |
4138 | * skb_get_timestamp - get timestamp from a skb | |
4139 | * @skb: skb to get stamp from | |
13c6ee2a | 4140 | * @stamp: pointer to struct __kernel_old_timeval to store stamp in |
a61bbcf2 PM |
4141 | * |
4142 | * Timestamps are stored in the skb as offsets to a base timestamp. | |
4143 | * This function converts the offset back to a struct timeval and stores | |
4144 | * it in stamp. | |
4145 | */ | |
ac45f602 | 4146 | static inline void skb_get_timestamp(const struct sk_buff *skb, |
13c6ee2a | 4147 | struct __kernel_old_timeval *stamp) |
a61bbcf2 | 4148 | { |
13c6ee2a | 4149 | *stamp = ns_to_kernel_old_timeval(skb->tstamp); |
a61bbcf2 PM |
4150 | } |
4151 | ||
887feae3 DD |
4152 | static inline void skb_get_new_timestamp(const struct sk_buff *skb, |
4153 | struct __kernel_sock_timeval *stamp) | |
4154 | { | |
4155 | struct timespec64 ts = ktime_to_timespec64(skb->tstamp); | |
4156 | ||
4157 | stamp->tv_sec = ts.tv_sec; | |
4158 | stamp->tv_usec = ts.tv_nsec / 1000; | |
4159 | } | |
4160 | ||
ac45f602 | 4161 | static inline void skb_get_timestampns(const struct sk_buff *skb, |
df1b4ba9 | 4162 | struct __kernel_old_timespec *stamp) |
ac45f602 | 4163 | { |
df1b4ba9 AB |
4164 | struct timespec64 ts = ktime_to_timespec64(skb->tstamp); |
4165 | ||
4166 | stamp->tv_sec = ts.tv_sec; | |
4167 | stamp->tv_nsec = ts.tv_nsec; | |
ac45f602 PO |
4168 | } |
4169 | ||
887feae3 DD |
4170 | static inline void skb_get_new_timestampns(const struct sk_buff *skb, |
4171 | struct __kernel_timespec *stamp) | |
4172 | { | |
4173 | struct timespec64 ts = ktime_to_timespec64(skb->tstamp); | |
4174 | ||
4175 | stamp->tv_sec = ts.tv_sec; | |
4176 | stamp->tv_nsec = ts.tv_nsec; | |
4177 | } | |
4178 | ||
b7aa0bf7 | 4179 | static inline void __net_timestamp(struct sk_buff *skb) |
a61bbcf2 | 4180 | { |
b7aa0bf7 | 4181 | skb->tstamp = ktime_get_real(); |
d93376f5 | 4182 | skb->mono_delivery_time = 0; |
a61bbcf2 PM |
4183 | } |
4184 | ||
164891aa SH |
4185 | static inline ktime_t net_timedelta(ktime_t t) |
4186 | { | |
4187 | return ktime_sub(ktime_get_real(), t); | |
4188 | } | |
4189 | ||
a1ac9c8a MKL |
4190 | static inline void skb_set_delivery_time(struct sk_buff *skb, ktime_t kt, |
4191 | bool mono) | |
4192 | { | |
4193 | skb->tstamp = kt; | |
d98d58a0 | 4194 | skb->mono_delivery_time = kt && mono; |
a1ac9c8a MKL |
4195 | } |
4196 | ||
27942a15 MKL |
4197 | DECLARE_STATIC_KEY_FALSE(netstamp_needed_key); |
4198 | ||
4199 | /* It is used in the ingress path to clear the delivery_time. | |
4200 | * If needed, set the skb->tstamp to the (rcv) timestamp. | |
4201 | */ | |
4202 | static inline void skb_clear_delivery_time(struct sk_buff *skb) | |
4203 | { | |
4204 | if (skb->mono_delivery_time) { | |
4205 | skb->mono_delivery_time = 0; | |
4206 | if (static_branch_unlikely(&netstamp_needed_key)) | |
4207 | skb->tstamp = ktime_get_real(); | |
4208 | else | |
4209 | skb->tstamp = 0; | |
4210 | } | |
4211 | } | |
4212 | ||
de799101 MKL |
4213 | static inline void skb_clear_tstamp(struct sk_buff *skb) |
4214 | { | |
4215 | if (skb->mono_delivery_time) | |
4216 | return; | |
4217 | ||
4218 | skb->tstamp = 0; | |
4219 | } | |
4220 | ||
27942a15 MKL |
4221 | static inline ktime_t skb_tstamp(const struct sk_buff *skb) |
4222 | { | |
4223 | if (skb->mono_delivery_time) | |
4224 | return 0; | |
4225 | ||
4226 | return skb->tstamp; | |
4227 | } | |
4228 | ||
b6561f84 MKL |
4229 | static inline ktime_t skb_tstamp_cond(const struct sk_buff *skb, bool cond) |
4230 | { | |
4231 | if (!skb->mono_delivery_time && skb->tstamp) | |
4232 | return skb->tstamp; | |
4233 | ||
4234 | if (static_branch_unlikely(&netstamp_needed_key) || cond) | |
4235 | return ktime_get_real(); | |
4236 | ||
4237 | return 0; | |
4238 | } | |
4239 | ||
de8f3a83 DB |
4240 | static inline u8 skb_metadata_len(const struct sk_buff *skb) |
4241 | { | |
4242 | return skb_shinfo(skb)->meta_len; | |
4243 | } | |
4244 | ||
4245 | static inline void *skb_metadata_end(const struct sk_buff *skb) | |
4246 | { | |
4247 | return skb_mac_header(skb); | |
4248 | } | |
4249 | ||
4250 | static inline bool __skb_metadata_differs(const struct sk_buff *skb_a, | |
4251 | const struct sk_buff *skb_b, | |
4252 | u8 meta_len) | |
4253 | { | |
4254 | const void *a = skb_metadata_end(skb_a); | |
4255 | const void *b = skb_metadata_end(skb_b); | |
de8f3a83 DB |
4256 | u64 diffs = 0; |
4257 | ||
2ebe81c8 AL |
4258 | if (!IS_ENABLED(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) || |
4259 | BITS_PER_LONG != 64) | |
4260 | goto slow; | |
4261 | ||
4262 | /* Using more efficient variant than plain call to memcmp(). */ | |
de8f3a83 DB |
4263 | switch (meta_len) { |
4264 | #define __it(x, op) (x -= sizeof(u##op)) | |
4265 | #define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op)) | |
4266 | case 32: diffs |= __it_diff(a, b, 64); | |
df561f66 | 4267 | fallthrough; |
de8f3a83 | 4268 | case 24: diffs |= __it_diff(a, b, 64); |
df561f66 | 4269 | fallthrough; |
de8f3a83 | 4270 | case 16: diffs |= __it_diff(a, b, 64); |
df561f66 | 4271 | fallthrough; |
de8f3a83 DB |
4272 | case 8: diffs |= __it_diff(a, b, 64); |
4273 | break; | |
4274 | case 28: diffs |= __it_diff(a, b, 64); | |
df561f66 | 4275 | fallthrough; |
de8f3a83 | 4276 | case 20: diffs |= __it_diff(a, b, 64); |
df561f66 | 4277 | fallthrough; |
de8f3a83 | 4278 | case 12: diffs |= __it_diff(a, b, 64); |
df561f66 | 4279 | fallthrough; |
de8f3a83 DB |
4280 | case 4: diffs |= __it_diff(a, b, 32); |
4281 | break; | |
2ebe81c8 AL |
4282 | default: |
4283 | slow: | |
4284 | return memcmp(a - meta_len, b - meta_len, meta_len); | |
de8f3a83 DB |
4285 | } |
4286 | return diffs; | |
de8f3a83 DB |
4287 | } |
4288 | ||
4289 | static inline bool skb_metadata_differs(const struct sk_buff *skb_a, | |
4290 | const struct sk_buff *skb_b) | |
4291 | { | |
4292 | u8 len_a = skb_metadata_len(skb_a); | |
4293 | u8 len_b = skb_metadata_len(skb_b); | |
4294 | ||
4295 | if (!(len_a | len_b)) | |
4296 | return false; | |
4297 | ||
4298 | return len_a != len_b ? | |
4299 | true : __skb_metadata_differs(skb_a, skb_b, len_a); | |
4300 | } | |
4301 | ||
4302 | static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len) | |
4303 | { | |
4304 | skb_shinfo(skb)->meta_len = meta_len; | |
4305 | } | |
4306 | ||
4307 | static inline void skb_metadata_clear(struct sk_buff *skb) | |
4308 | { | |
4309 | skb_metadata_set(skb, 0); | |
4310 | } | |
4311 | ||
62bccb8c AD |
4312 | struct sk_buff *skb_clone_sk(struct sk_buff *skb); |
4313 | ||
c1f19b51 RC |
4314 | #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING |
4315 | ||
7965bd4d JP |
4316 | void skb_clone_tx_timestamp(struct sk_buff *skb); |
4317 | bool skb_defer_rx_timestamp(struct sk_buff *skb); | |
c1f19b51 RC |
4318 | |
4319 | #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */ | |
4320 | ||
4321 | static inline void skb_clone_tx_timestamp(struct sk_buff *skb) | |
4322 | { | |
4323 | } | |
4324 | ||
4325 | static inline bool skb_defer_rx_timestamp(struct sk_buff *skb) | |
4326 | { | |
4327 | return false; | |
4328 | } | |
4329 | ||
4330 | #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */ | |
4331 | ||
4332 | /** | |
4333 | * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps | |
4334 | * | |
da92b194 RC |
4335 | * PHY drivers may accept clones of transmitted packets for |
4336 | * timestamping via their phy_driver.txtstamp method. These drivers | |
7a76a021 BP |
4337 | * must call this function to return the skb back to the stack with a |
4338 | * timestamp. | |
da92b194 | 4339 | * |
2ff17117 | 4340 | * @skb: clone of the original outgoing packet |
7a76a021 | 4341 | * @hwtstamps: hardware time stamps |
c1f19b51 RC |
4342 | * |
4343 | */ | |
4344 | void skb_complete_tx_timestamp(struct sk_buff *skb, | |
4345 | struct skb_shared_hwtstamps *hwtstamps); | |
4346 | ||
e7ed11ee | 4347 | void __skb_tstamp_tx(struct sk_buff *orig_skb, const struct sk_buff *ack_skb, |
e7fd2885 WB |
4348 | struct skb_shared_hwtstamps *hwtstamps, |
4349 | struct sock *sk, int tstype); | |
4350 | ||
ac45f602 PO |
4351 | /** |
4352 | * skb_tstamp_tx - queue clone of skb with send time stamps | |
4353 | * @orig_skb: the original outgoing packet | |
4354 | * @hwtstamps: hardware time stamps, may be NULL if not available | |
4355 | * | |
4356 | * If the skb has a socket associated, then this function clones the | |
4357 | * skb (thus sharing the actual data and optional structures), stores | |
4358 | * the optional hardware time stamping information (if non NULL) or | |
4359 | * generates a software time stamp (otherwise), then queues the clone | |
4360 | * to the error queue of the socket. Errors are silently ignored. | |
4361 | */ | |
7965bd4d JP |
4362 | void skb_tstamp_tx(struct sk_buff *orig_skb, |
4363 | struct skb_shared_hwtstamps *hwtstamps); | |
ac45f602 | 4364 | |
4507a715 RC |
4365 | /** |
4366 | * skb_tx_timestamp() - Driver hook for transmit timestamping | |
4367 | * | |
4368 | * Ethernet MAC Drivers should call this function in their hard_xmit() | |
4ff75b7c | 4369 | * function immediately before giving the sk_buff to the MAC hardware. |
4507a715 | 4370 | * |
73409f3b DM |
4371 | * Specifically, one should make absolutely sure that this function is |
4372 | * called before TX completion of this packet can trigger. Otherwise | |
4373 | * the packet could potentially already be freed. | |
4374 | * | |
4507a715 RC |
4375 | * @skb: A socket buffer. |
4376 | */ | |
4377 | static inline void skb_tx_timestamp(struct sk_buff *skb) | |
4378 | { | |
c1f19b51 | 4379 | skb_clone_tx_timestamp(skb); |
b50a5c70 ML |
4380 | if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP) |
4381 | skb_tstamp_tx(skb, NULL); | |
4507a715 RC |
4382 | } |
4383 | ||
6e3e939f JB |
4384 | /** |
4385 | * skb_complete_wifi_ack - deliver skb with wifi status | |
4386 | * | |
4387 | * @skb: the original outgoing packet | |
4388 | * @acked: ack status | |
4389 | * | |
4390 | */ | |
4391 | void skb_complete_wifi_ack(struct sk_buff *skb, bool acked); | |
4392 | ||
7965bd4d JP |
4393 | __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len); |
4394 | __sum16 __skb_checksum_complete(struct sk_buff *skb); | |
fb286bb2 | 4395 | |
60476372 HX |
4396 | static inline int skb_csum_unnecessary(const struct sk_buff *skb) |
4397 | { | |
6edec0e6 TH |
4398 | return ((skb->ip_summed == CHECKSUM_UNNECESSARY) || |
4399 | skb->csum_valid || | |
4400 | (skb->ip_summed == CHECKSUM_PARTIAL && | |
4401 | skb_checksum_start_offset(skb) >= 0)); | |
60476372 HX |
4402 | } |
4403 | ||
fb286bb2 HX |
4404 | /** |
4405 | * skb_checksum_complete - Calculate checksum of an entire packet | |
4406 | * @skb: packet to process | |
4407 | * | |
4408 | * This function calculates the checksum over the entire packet plus | |
4409 | * the value of skb->csum. The latter can be used to supply the | |
4410 | * checksum of a pseudo header as used by TCP/UDP. It returns the | |
4411 | * checksum. | |
4412 | * | |
4413 | * For protocols that contain complete checksums such as ICMP/TCP/UDP, | |
4414 | * this function can be used to verify that checksum on received | |
4415 | * packets. In that case the function should return zero if the | |
4416 | * checksum is correct. In particular, this function will return zero | |
4417 | * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the | |
4418 | * hardware has already verified the correctness of the checksum. | |
4419 | */ | |
4381ca3c | 4420 | static inline __sum16 skb_checksum_complete(struct sk_buff *skb) |
fb286bb2 | 4421 | { |
60476372 HX |
4422 | return skb_csum_unnecessary(skb) ? |
4423 | 0 : __skb_checksum_complete(skb); | |
fb286bb2 HX |
4424 | } |
4425 | ||
77cffe23 TH |
4426 | static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb) |
4427 | { | |
4428 | if (skb->ip_summed == CHECKSUM_UNNECESSARY) { | |
4429 | if (skb->csum_level == 0) | |
4430 | skb->ip_summed = CHECKSUM_NONE; | |
4431 | else | |
4432 | skb->csum_level--; | |
4433 | } | |
4434 | } | |
4435 | ||
4436 | static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb) | |
4437 | { | |
4438 | if (skb->ip_summed == CHECKSUM_UNNECESSARY) { | |
4439 | if (skb->csum_level < SKB_MAX_CSUM_LEVEL) | |
4440 | skb->csum_level++; | |
4441 | } else if (skb->ip_summed == CHECKSUM_NONE) { | |
4442 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
4443 | skb->csum_level = 0; | |
4444 | } | |
4445 | } | |
4446 | ||
836e66c2 DB |
4447 | static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb) |
4448 | { | |
4449 | if (skb->ip_summed == CHECKSUM_UNNECESSARY) { | |
4450 | skb->ip_summed = CHECKSUM_NONE; | |
4451 | skb->csum_level = 0; | |
4452 | } | |
4453 | } | |
4454 | ||
76ba0aae TH |
4455 | /* Check if we need to perform checksum complete validation. |
4456 | * | |
4457 | * Returns true if checksum complete is needed, false otherwise | |
4458 | * (either checksum is unnecessary or zero checksum is allowed). | |
4459 | */ | |
4460 | static inline bool __skb_checksum_validate_needed(struct sk_buff *skb, | |
4461 | bool zero_okay, | |
4462 | __sum16 check) | |
4463 | { | |
5d0c2b95 TH |
4464 | if (skb_csum_unnecessary(skb) || (zero_okay && !check)) { |
4465 | skb->csum_valid = 1; | |
77cffe23 | 4466 | __skb_decr_checksum_unnecessary(skb); |
76ba0aae TH |
4467 | return false; |
4468 | } | |
4469 | ||
4470 | return true; | |
4471 | } | |
4472 | ||
da279887 | 4473 | /* For small packets <= CHECKSUM_BREAK perform checksum complete directly |
76ba0aae TH |
4474 | * in checksum_init. |
4475 | */ | |
4476 | #define CHECKSUM_BREAK 76 | |
4477 | ||
4e18b9ad TH |
4478 | /* Unset checksum-complete |
4479 | * | |
4480 | * Unset checksum complete can be done when packet is being modified | |
4481 | * (uncompressed for instance) and checksum-complete value is | |
4482 | * invalidated. | |
4483 | */ | |
4484 | static inline void skb_checksum_complete_unset(struct sk_buff *skb) | |
4485 | { | |
4486 | if (skb->ip_summed == CHECKSUM_COMPLETE) | |
4487 | skb->ip_summed = CHECKSUM_NONE; | |
4488 | } | |
4489 | ||
76ba0aae TH |
4490 | /* Validate (init) checksum based on checksum complete. |
4491 | * | |
4492 | * Return values: | |
4493 | * 0: checksum is validated or try to in skb_checksum_complete. In the latter | |
4494 | * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo | |
4495 | * checksum is stored in skb->csum for use in __skb_checksum_complete | |
4496 | * non-zero: value of invalid checksum | |
4497 | * | |
4498 | */ | |
4499 | static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb, | |
4500 | bool complete, | |
4501 | __wsum psum) | |
4502 | { | |
4503 | if (skb->ip_summed == CHECKSUM_COMPLETE) { | |
4504 | if (!csum_fold(csum_add(psum, skb->csum))) { | |
5d0c2b95 | 4505 | skb->csum_valid = 1; |
76ba0aae TH |
4506 | return 0; |
4507 | } | |
4508 | } | |
4509 | ||
4510 | skb->csum = psum; | |
4511 | ||
5d0c2b95 TH |
4512 | if (complete || skb->len <= CHECKSUM_BREAK) { |
4513 | __sum16 csum; | |
4514 | ||
4515 | csum = __skb_checksum_complete(skb); | |
4516 | skb->csum_valid = !csum; | |
4517 | return csum; | |
4518 | } | |
76ba0aae TH |
4519 | |
4520 | return 0; | |
4521 | } | |
4522 | ||
4523 | static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto) | |
4524 | { | |
4525 | return 0; | |
4526 | } | |
4527 | ||
4528 | /* Perform checksum validate (init). Note that this is a macro since we only | |
4529 | * want to calculate the pseudo header which is an input function if necessary. | |
4530 | * First we try to validate without any computation (checksum unnecessary) and | |
4531 | * then calculate based on checksum complete calling the function to compute | |
4532 | * pseudo header. | |
4533 | * | |
4534 | * Return values: | |
4535 | * 0: checksum is validated or try to in skb_checksum_complete | |
4536 | * non-zero: value of invalid checksum | |
4537 | */ | |
4538 | #define __skb_checksum_validate(skb, proto, complete, \ | |
4539 | zero_okay, check, compute_pseudo) \ | |
4540 | ({ \ | |
4541 | __sum16 __ret = 0; \ | |
5d0c2b95 | 4542 | skb->csum_valid = 0; \ |
76ba0aae TH |
4543 | if (__skb_checksum_validate_needed(skb, zero_okay, check)) \ |
4544 | __ret = __skb_checksum_validate_complete(skb, \ | |
4545 | complete, compute_pseudo(skb, proto)); \ | |
4546 | __ret; \ | |
4547 | }) | |
4548 | ||
4549 | #define skb_checksum_init(skb, proto, compute_pseudo) \ | |
4550 | __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo) | |
4551 | ||
4552 | #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \ | |
4553 | __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo) | |
4554 | ||
4555 | #define skb_checksum_validate(skb, proto, compute_pseudo) \ | |
4556 | __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo) | |
4557 | ||
4558 | #define skb_checksum_validate_zero_check(skb, proto, check, \ | |
4559 | compute_pseudo) \ | |
096a4cfa | 4560 | __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo) |
76ba0aae TH |
4561 | |
4562 | #define skb_checksum_simple_validate(skb) \ | |
4563 | __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo) | |
4564 | ||
d96535a1 TH |
4565 | static inline bool __skb_checksum_convert_check(struct sk_buff *skb) |
4566 | { | |
219f1d79 | 4567 | return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid); |
d96535a1 TH |
4568 | } |
4569 | ||
e4aa33ad | 4570 | static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo) |
d96535a1 TH |
4571 | { |
4572 | skb->csum = ~pseudo; | |
4573 | skb->ip_summed = CHECKSUM_COMPLETE; | |
4574 | } | |
4575 | ||
e4aa33ad | 4576 | #define skb_checksum_try_convert(skb, proto, compute_pseudo) \ |
d96535a1 TH |
4577 | do { \ |
4578 | if (__skb_checksum_convert_check(skb)) \ | |
e4aa33ad | 4579 | __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \ |
d96535a1 TH |
4580 | } while (0) |
4581 | ||
15e2396d TH |
4582 | static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr, |
4583 | u16 start, u16 offset) | |
4584 | { | |
4585 | skb->ip_summed = CHECKSUM_PARTIAL; | |
4586 | skb->csum_start = ((unsigned char *)ptr + start) - skb->head; | |
4587 | skb->csum_offset = offset - start; | |
4588 | } | |
4589 | ||
dcdc8994 TH |
4590 | /* Update skbuf and packet to reflect the remote checksum offload operation. |
4591 | * When called, ptr indicates the starting point for skb->csum when | |
4592 | * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete | |
4593 | * here, skb_postpull_rcsum is done so skb->csum start is ptr. | |
4594 | */ | |
4595 | static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr, | |
15e2396d | 4596 | int start, int offset, bool nopartial) |
dcdc8994 TH |
4597 | { |
4598 | __wsum delta; | |
4599 | ||
15e2396d TH |
4600 | if (!nopartial) { |
4601 | skb_remcsum_adjust_partial(skb, ptr, start, offset); | |
4602 | return; | |
4603 | } | |
4604 | ||
10a2308f | 4605 | if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) { |
dcdc8994 TH |
4606 | __skb_checksum_complete(skb); |
4607 | skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data); | |
4608 | } | |
4609 | ||
4610 | delta = remcsum_adjust(ptr, skb->csum, start, offset); | |
4611 | ||
4612 | /* Adjust skb->csum since we changed the packet */ | |
4613 | skb->csum = csum_add(skb->csum, delta); | |
4614 | } | |
4615 | ||
cb9c6836 FW |
4616 | static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb) |
4617 | { | |
4618 | #if IS_ENABLED(CONFIG_NF_CONNTRACK) | |
261db6c2 | 4619 | return (void *)(skb->_nfct & NFCT_PTRMASK); |
cb9c6836 FW |
4620 | #else |
4621 | return NULL; | |
4622 | #endif | |
4623 | } | |
4624 | ||
261db6c2 | 4625 | static inline unsigned long skb_get_nfct(const struct sk_buff *skb) |
1da177e4 | 4626 | { |
261db6c2 JS |
4627 | #if IS_ENABLED(CONFIG_NF_CONNTRACK) |
4628 | return skb->_nfct; | |
4629 | #else | |
4630 | return 0UL; | |
4631 | #endif | |
1da177e4 | 4632 | } |
261db6c2 JS |
4633 | |
4634 | static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct) | |
1da177e4 | 4635 | { |
261db6c2 | 4636 | #if IS_ENABLED(CONFIG_NF_CONNTRACK) |
5fc88f93 | 4637 | skb->slow_gro |= !!nfct; |
261db6c2 | 4638 | skb->_nfct = nfct; |
2fc72c7b | 4639 | #endif |
261db6c2 | 4640 | } |
df5042f4 FW |
4641 | |
4642 | #ifdef CONFIG_SKB_EXTENSIONS | |
4643 | enum skb_ext_id { | |
4644 | #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) | |
4645 | SKB_EXT_BRIDGE_NF, | |
4165079b FW |
4646 | #endif |
4647 | #ifdef CONFIG_XFRM | |
4648 | SKB_EXT_SEC_PATH, | |
95a7233c PB |
4649 | #endif |
4650 | #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT) | |
4651 | TC_SKB_EXT, | |
3ee17bc7 MM |
4652 | #endif |
4653 | #if IS_ENABLED(CONFIG_MPTCP) | |
4654 | SKB_EXT_MPTCP, | |
78476d31 JK |
4655 | #endif |
4656 | #if IS_ENABLED(CONFIG_MCTP_FLOWS) | |
4657 | SKB_EXT_MCTP, | |
df5042f4 FW |
4658 | #endif |
4659 | SKB_EXT_NUM, /* must be last */ | |
4660 | }; | |
4661 | ||
4662 | /** | |
4663 | * struct skb_ext - sk_buff extensions | |
4664 | * @refcnt: 1 on allocation, deallocated on 0 | |
4665 | * @offset: offset to add to @data to obtain extension address | |
4666 | * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units | |
4667 | * @data: start of extension data, variable sized | |
4668 | * | |
4669 | * Note: offsets/lengths are stored in chunks of 8 bytes, this allows | |
4670 | * to use 'u8' types while allowing up to 2kb worth of extension data. | |
4671 | */ | |
4672 | struct skb_ext { | |
4673 | refcount_t refcnt; | |
4674 | u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */ | |
4675 | u8 chunks; /* same */ | |
5c91aa1d | 4676 | char data[] __aligned(8); |
df5042f4 FW |
4677 | }; |
4678 | ||
4930f483 | 4679 | struct skb_ext *__skb_ext_alloc(gfp_t flags); |
8b69a803 PA |
4680 | void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id, |
4681 | struct skb_ext *ext); | |
df5042f4 FW |
4682 | void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id); |
4683 | void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id); | |
4684 | void __skb_ext_put(struct skb_ext *ext); | |
4685 | ||
4686 | static inline void skb_ext_put(struct sk_buff *skb) | |
4687 | { | |
4688 | if (skb->active_extensions) | |
4689 | __skb_ext_put(skb->extensions); | |
4690 | } | |
4691 | ||
df5042f4 FW |
4692 | static inline void __skb_ext_copy(struct sk_buff *dst, |
4693 | const struct sk_buff *src) | |
4694 | { | |
4695 | dst->active_extensions = src->active_extensions; | |
4696 | ||
4697 | if (src->active_extensions) { | |
4698 | struct skb_ext *ext = src->extensions; | |
4699 | ||
4700 | refcount_inc(&ext->refcnt); | |
4701 | dst->extensions = ext; | |
4702 | } | |
4703 | } | |
4704 | ||
4705 | static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src) | |
4706 | { | |
4707 | skb_ext_put(dst); | |
4708 | __skb_ext_copy(dst, src); | |
4709 | } | |
4710 | ||
4711 | static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i) | |
4712 | { | |
4713 | return !!ext->offset[i]; | |
4714 | } | |
4715 | ||
4716 | static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id) | |
4717 | { | |
4718 | return skb->active_extensions & (1 << id); | |
4719 | } | |
4720 | ||
4721 | static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id) | |
4722 | { | |
4723 | if (skb_ext_exist(skb, id)) | |
4724 | __skb_ext_del(skb, id); | |
4725 | } | |
4726 | ||
4727 | static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id) | |
4728 | { | |
4729 | if (skb_ext_exist(skb, id)) { | |
4730 | struct skb_ext *ext = skb->extensions; | |
4731 | ||
4732 | return (void *)ext + (ext->offset[id] << 3); | |
4733 | } | |
4734 | ||
4735 | return NULL; | |
4736 | } | |
174e2381 FW |
4737 | |
4738 | static inline void skb_ext_reset(struct sk_buff *skb) | |
4739 | { | |
4740 | if (unlikely(skb->active_extensions)) { | |
4741 | __skb_ext_put(skb->extensions); | |
4742 | skb->active_extensions = 0; | |
4743 | } | |
4744 | } | |
677bf08c FW |
4745 | |
4746 | static inline bool skb_has_extensions(struct sk_buff *skb) | |
4747 | { | |
4748 | return unlikely(skb->active_extensions); | |
4749 | } | |
df5042f4 FW |
4750 | #else |
4751 | static inline void skb_ext_put(struct sk_buff *skb) {} | |
174e2381 | 4752 | static inline void skb_ext_reset(struct sk_buff *skb) {} |
df5042f4 FW |
4753 | static inline void skb_ext_del(struct sk_buff *skb, int unused) {} |
4754 | static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {} | |
4755 | static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {} | |
677bf08c | 4756 | static inline bool skb_has_extensions(struct sk_buff *skb) { return false; } |
df5042f4 FW |
4757 | #endif /* CONFIG_SKB_EXTENSIONS */ |
4758 | ||
895b5c9f | 4759 | static inline void nf_reset_ct(struct sk_buff *skb) |
a193a4ab | 4760 | { |
5f79e0f9 | 4761 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) |
a9e419dc FW |
4762 | nf_conntrack_put(skb_nfct(skb)); |
4763 | skb->_nfct = 0; | |
2fc72c7b | 4764 | #endif |
a193a4ab PM |
4765 | } |
4766 | ||
124dff01 PM |
4767 | static inline void nf_reset_trace(struct sk_buff *skb) |
4768 | { | |
c55c0e91 | 4769 | #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || IS_ENABLED(CONFIG_NF_TABLES) |
130549fe G |
4770 | skb->nf_trace = 0; |
4771 | #endif | |
a193a4ab PM |
4772 | } |
4773 | ||
2b5ec1a5 YY |
4774 | static inline void ipvs_reset(struct sk_buff *skb) |
4775 | { | |
4776 | #if IS_ENABLED(CONFIG_IP_VS) | |
4777 | skb->ipvs_property = 0; | |
4778 | #endif | |
4779 | } | |
4780 | ||
de8bda1d | 4781 | /* Note: This doesn't put any conntrack info in dst. */ |
b1937227 ED |
4782 | static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src, |
4783 | bool copy) | |
edda553c | 4784 | { |
5f79e0f9 | 4785 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) |
a9e419dc FW |
4786 | dst->_nfct = src->_nfct; |
4787 | nf_conntrack_get(skb_nfct(src)); | |
2fc72c7b | 4788 | #endif |
c55c0e91 | 4789 | #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || IS_ENABLED(CONFIG_NF_TABLES) |
b1937227 ED |
4790 | if (copy) |
4791 | dst->nf_trace = src->nf_trace; | |
478b360a | 4792 | #endif |
edda553c YK |
4793 | } |
4794 | ||
e7ac05f3 YK |
4795 | static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src) |
4796 | { | |
e7ac05f3 | 4797 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) |
a9e419dc | 4798 | nf_conntrack_put(skb_nfct(dst)); |
e7ac05f3 | 4799 | #endif |
5fc88f93 | 4800 | dst->slow_gro = src->slow_gro; |
b1937227 | 4801 | __nf_copy(dst, src, true); |
e7ac05f3 YK |
4802 | } |
4803 | ||
984bc16c JM |
4804 | #ifdef CONFIG_NETWORK_SECMARK |
4805 | static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) | |
4806 | { | |
4807 | to->secmark = from->secmark; | |
4808 | } | |
4809 | ||
4810 | static inline void skb_init_secmark(struct sk_buff *skb) | |
4811 | { | |
4812 | skb->secmark = 0; | |
4813 | } | |
4814 | #else | |
4815 | static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) | |
4816 | { } | |
4817 | ||
4818 | static inline void skb_init_secmark(struct sk_buff *skb) | |
4819 | { } | |
4820 | #endif | |
4821 | ||
7af8f4ca FW |
4822 | static inline int secpath_exists(const struct sk_buff *skb) |
4823 | { | |
4824 | #ifdef CONFIG_XFRM | |
4165079b | 4825 | return skb_ext_exist(skb, SKB_EXT_SEC_PATH); |
7af8f4ca FW |
4826 | #else |
4827 | return 0; | |
4828 | #endif | |
4829 | } | |
4830 | ||
574f7194 EB |
4831 | static inline bool skb_irq_freeable(const struct sk_buff *skb) |
4832 | { | |
4833 | return !skb->destructor && | |
7af8f4ca | 4834 | !secpath_exists(skb) && |
cb9c6836 | 4835 | !skb_nfct(skb) && |
574f7194 EB |
4836 | !skb->_skb_refdst && |
4837 | !skb_has_frag_list(skb); | |
4838 | } | |
4839 | ||
f25f4e44 PWJ |
4840 | static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping) |
4841 | { | |
f25f4e44 | 4842 | skb->queue_mapping = queue_mapping; |
f25f4e44 PWJ |
4843 | } |
4844 | ||
9247744e | 4845 | static inline u16 skb_get_queue_mapping(const struct sk_buff *skb) |
4e3ab47a | 4846 | { |
4e3ab47a | 4847 | return skb->queue_mapping; |
4e3ab47a PE |
4848 | } |
4849 | ||
f25f4e44 PWJ |
4850 | static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from) |
4851 | { | |
f25f4e44 | 4852 | to->queue_mapping = from->queue_mapping; |
f25f4e44 PWJ |
4853 | } |
4854 | ||
d5a9e24a DM |
4855 | static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue) |
4856 | { | |
4857 | skb->queue_mapping = rx_queue + 1; | |
4858 | } | |
4859 | ||
9247744e | 4860 | static inline u16 skb_get_rx_queue(const struct sk_buff *skb) |
d5a9e24a DM |
4861 | { |
4862 | return skb->queue_mapping - 1; | |
4863 | } | |
4864 | ||
9247744e | 4865 | static inline bool skb_rx_queue_recorded(const struct sk_buff *skb) |
d5a9e24a | 4866 | { |
a02cec21 | 4867 | return skb->queue_mapping != 0; |
d5a9e24a DM |
4868 | } |
4869 | ||
4ff06203 JA |
4870 | static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val) |
4871 | { | |
4872 | skb->dst_pending_confirm = val; | |
4873 | } | |
4874 | ||
4875 | static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb) | |
4876 | { | |
4877 | return skb->dst_pending_confirm != 0; | |
4878 | } | |
4879 | ||
2294be0f | 4880 | static inline struct sec_path *skb_sec_path(const struct sk_buff *skb) |
def8b4fa | 4881 | { |
0b3d8e08 | 4882 | #ifdef CONFIG_XFRM |
4165079b | 4883 | return skb_ext_find(skb, SKB_EXT_SEC_PATH); |
def8b4fa | 4884 | #else |
def8b4fa | 4885 | return NULL; |
def8b4fa | 4886 | #endif |
0b3d8e08 | 4887 | } |
def8b4fa | 4888 | |
bdcc0924 | 4889 | static inline bool skb_is_gso(const struct sk_buff *skb) |
89114afd HX |
4890 | { |
4891 | return skb_shinfo(skb)->gso_size; | |
4892 | } | |
4893 | ||
36a8f39e | 4894 | /* Note: Should be called only if skb_is_gso(skb) is true */ |
bdcc0924 | 4895 | static inline bool skb_is_gso_v6(const struct sk_buff *skb) |
eabd7e35 BG |
4896 | { |
4897 | return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6; | |
4898 | } | |
4899 | ||
d02f51cb DA |
4900 | /* Note: Should be called only if skb_is_gso(skb) is true */ |
4901 | static inline bool skb_is_gso_sctp(const struct sk_buff *skb) | |
4902 | { | |
4903 | return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP; | |
4904 | } | |
4905 | ||
4c3024de | 4906 | /* Note: Should be called only if skb_is_gso(skb) is true */ |
b90efd22 WB |
4907 | static inline bool skb_is_gso_tcp(const struct sk_buff *skb) |
4908 | { | |
4c3024de | 4909 | return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6); |
b90efd22 WB |
4910 | } |
4911 | ||
5293efe6 DB |
4912 | static inline void skb_gso_reset(struct sk_buff *skb) |
4913 | { | |
4914 | skb_shinfo(skb)->gso_size = 0; | |
4915 | skb_shinfo(skb)->gso_segs = 0; | |
4916 | skb_shinfo(skb)->gso_type = 0; | |
4917 | } | |
4918 | ||
d02f51cb DA |
4919 | static inline void skb_increase_gso_size(struct skb_shared_info *shinfo, |
4920 | u16 increment) | |
4921 | { | |
4922 | if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS)) | |
4923 | return; | |
4924 | shinfo->gso_size += increment; | |
4925 | } | |
4926 | ||
4927 | static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo, | |
4928 | u16 decrement) | |
4929 | { | |
4930 | if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS)) | |
4931 | return; | |
4932 | shinfo->gso_size -= decrement; | |
4933 | } | |
4934 | ||
7965bd4d | 4935 | void __skb_warn_lro_forwarding(const struct sk_buff *skb); |
4497b076 BH |
4936 | |
4937 | static inline bool skb_warn_if_lro(const struct sk_buff *skb) | |
4938 | { | |
4939 | /* LRO sets gso_size but not gso_type, whereas if GSO is really | |
4940 | * wanted then gso_type will be set. */ | |
05bdd2f1 ED |
4941 | const struct skb_shared_info *shinfo = skb_shinfo(skb); |
4942 | ||
b78462eb AD |
4943 | if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 && |
4944 | unlikely(shinfo->gso_type == 0)) { | |
4497b076 BH |
4945 | __skb_warn_lro_forwarding(skb); |
4946 | return true; | |
4947 | } | |
4948 | return false; | |
4949 | } | |
4950 | ||
35fc92a9 HX |
4951 | static inline void skb_forward_csum(struct sk_buff *skb) |
4952 | { | |
4953 | /* Unfortunately we don't support this one. Any brave souls? */ | |
4954 | if (skb->ip_summed == CHECKSUM_COMPLETE) | |
4955 | skb->ip_summed = CHECKSUM_NONE; | |
4956 | } | |
4957 | ||
bc8acf2c ED |
4958 | /** |
4959 | * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE | |
4960 | * @skb: skb to check | |
4961 | * | |
4962 | * fresh skbs have their ip_summed set to CHECKSUM_NONE. | |
4963 | * Instead of forcing ip_summed to CHECKSUM_NONE, we can | |
4964 | * use this helper, to document places where we make this assertion. | |
4965 | */ | |
05bdd2f1 | 4966 | static inline void skb_checksum_none_assert(const struct sk_buff *skb) |
bc8acf2c | 4967 | { |
0df65743 | 4968 | DEBUG_NET_WARN_ON_ONCE(skb->ip_summed != CHECKSUM_NONE); |
bc8acf2c ED |
4969 | } |
4970 | ||
f35d9d8a | 4971 | bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off); |
a6686f2f | 4972 | |
ed1f50c3 | 4973 | int skb_checksum_setup(struct sk_buff *skb, bool recalculate); |
9afd85c9 LL |
4974 | struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb, |
4975 | unsigned int transport_len, | |
4976 | __sum16(*skb_chkf)(struct sk_buff *skb)); | |
ed1f50c3 | 4977 | |
3a7c1ee4 AD |
4978 | /** |
4979 | * skb_head_is_locked - Determine if the skb->head is locked down | |
4980 | * @skb: skb to check | |
4981 | * | |
4982 | * The head on skbs build around a head frag can be removed if they are | |
4983 | * not cloned. This function returns true if the skb head is locked down | |
4984 | * due to either being allocated via kmalloc, or by being a clone with | |
4985 | * multiple references to the head. | |
4986 | */ | |
4987 | static inline bool skb_head_is_locked(const struct sk_buff *skb) | |
4988 | { | |
4989 | return !skb->head_frag || skb_cloned(skb); | |
4990 | } | |
fe6cc55f | 4991 | |
179bc67f EC |
4992 | /* Local Checksum Offload. |
4993 | * Compute outer checksum based on the assumption that the | |
4994 | * inner checksum will be offloaded later. | |
d0dcde64 | 4995 | * See Documentation/networking/checksum-offloads.rst for |
e8ae7b00 | 4996 | * explanation of how this works. |
179bc67f EC |
4997 | * Fill in outer checksum adjustment (e.g. with sum of outer |
4998 | * pseudo-header) before calling. | |
4999 | * Also ensure that inner checksum is in linear data area. | |
5000 | */ | |
5001 | static inline __wsum lco_csum(struct sk_buff *skb) | |
5002 | { | |
9e74a6da AD |
5003 | unsigned char *csum_start = skb_checksum_start(skb); |
5004 | unsigned char *l4_hdr = skb_transport_header(skb); | |
5005 | __wsum partial; | |
179bc67f EC |
5006 | |
5007 | /* Start with complement of inner checksum adjustment */ | |
9e74a6da AD |
5008 | partial = ~csum_unfold(*(__force __sum16 *)(csum_start + |
5009 | skb->csum_offset)); | |
5010 | ||
179bc67f | 5011 | /* Add in checksum of our headers (incl. outer checksum |
9e74a6da | 5012 | * adjustment filled in by caller) and return result. |
179bc67f | 5013 | */ |
9e74a6da | 5014 | return csum_partial(l4_hdr, csum_start - l4_hdr, partial); |
179bc67f EC |
5015 | } |
5016 | ||
2c64605b PNA |
5017 | static inline bool skb_is_redirected(const struct sk_buff *skb) |
5018 | { | |
2c64605b | 5019 | return skb->redirected; |
2c64605b PNA |
5020 | } |
5021 | ||
5022 | static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress) | |
5023 | { | |
2c64605b | 5024 | skb->redirected = 1; |
11941f8a | 5025 | #ifdef CONFIG_NET_REDIRECT |
2c64605b PNA |
5026 | skb->from_ingress = from_ingress; |
5027 | if (skb->from_ingress) | |
de799101 | 5028 | skb_clear_tstamp(skb); |
2c64605b PNA |
5029 | #endif |
5030 | } | |
5031 | ||
5032 | static inline void skb_reset_redirect(struct sk_buff *skb) | |
5033 | { | |
2c64605b | 5034 | skb->redirected = 0; |
2c64605b PNA |
5035 | } |
5036 | ||
59e498a3 DB |
5037 | static inline void skb_set_redirected_noclear(struct sk_buff *skb, |
5038 | bool from_ingress) | |
5039 | { | |
5040 | skb->redirected = 1; | |
5041 | #ifdef CONFIG_NET_REDIRECT | |
5042 | skb->from_ingress = from_ingress; | |
5043 | #endif | |
5044 | } | |
5045 | ||
fa821170 XL |
5046 | static inline bool skb_csum_is_sctp(struct sk_buff *skb) |
5047 | { | |
c24831a1 | 5048 | #if IS_ENABLED(CONFIG_IP_SCTP) |
fa821170 | 5049 | return skb->csum_not_inet; |
c24831a1 JK |
5050 | #else |
5051 | return 0; | |
5052 | #endif | |
5053 | } | |
5054 | ||
5055 | static inline void skb_reset_csum_not_inet(struct sk_buff *skb) | |
5056 | { | |
5057 | skb->ip_summed = CHECKSUM_NONE; | |
5058 | #if IS_ENABLED(CONFIG_IP_SCTP) | |
5059 | skb->csum_not_inet = 0; | |
5060 | #endif | |
fa821170 XL |
5061 | } |
5062 | ||
6370cc3b AN |
5063 | static inline void skb_set_kcov_handle(struct sk_buff *skb, |
5064 | const u64 kcov_handle) | |
5065 | { | |
fa69ee5a ME |
5066 | #ifdef CONFIG_KCOV |
5067 | skb->kcov_handle = kcov_handle; | |
5068 | #endif | |
6370cc3b AN |
5069 | } |
5070 | ||
5071 | static inline u64 skb_get_kcov_handle(struct sk_buff *skb) | |
5072 | { | |
fa69ee5a ME |
5073 | #ifdef CONFIG_KCOV |
5074 | return skb->kcov_handle; | |
6370cc3b | 5075 | #else |
fa69ee5a ME |
5076 | return 0; |
5077 | #endif | |
5078 | } | |
6370cc3b | 5079 | |
57f05bc2 | 5080 | static inline void skb_mark_for_recycle(struct sk_buff *skb) |
6a5bcd84 | 5081 | { |
2c854e5f | 5082 | #ifdef CONFIG_PAGE_POOL |
6a5bcd84 | 5083 | skb->pp_recycle = 1; |
6a5bcd84 | 5084 | #endif |
2c854e5f | 5085 | } |
6a5bcd84 | 5086 | |
2e910b95 DH |
5087 | ssize_t skb_splice_from_iter(struct sk_buff *skb, struct iov_iter *iter, |
5088 | ssize_t maxsize, gfp_t gfp); | |
5089 | ||
1da177e4 LT |
5090 | #endif /* __KERNEL__ */ |
5091 | #endif /* _LINUX_SKBUFF_H */ |