Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * Definitions for the 'struct sk_buff' memory handlers. | |
3 | * | |
4 | * Authors: | |
5 | * Alan Cox, <gw4pts@gw4pts.ampr.org> | |
6 | * Florian La Roche, <rzsfl@rz.uni-sb.de> | |
7 | * | |
8 | * This program is free software; you can redistribute it and/or | |
9 | * modify it under the terms of the GNU General Public License | |
10 | * as published by the Free Software Foundation; either version | |
11 | * 2 of the License, or (at your option) any later version. | |
12 | */ | |
13 | ||
14 | #ifndef _LINUX_SKBUFF_H | |
15 | #define _LINUX_SKBUFF_H | |
16 | ||
1da177e4 | 17 | #include <linux/kernel.h> |
fe55f6d5 | 18 | #include <linux/kmemcheck.h> |
1da177e4 LT |
19 | #include <linux/compiler.h> |
20 | #include <linux/time.h> | |
187f1882 | 21 | #include <linux/bug.h> |
1da177e4 | 22 | #include <linux/cache.h> |
56b17425 | 23 | #include <linux/rbtree.h> |
51f3d02b | 24 | #include <linux/socket.h> |
1da177e4 | 25 | |
60063497 | 26 | #include <linux/atomic.h> |
1da177e4 LT |
27 | #include <asm/types.h> |
28 | #include <linux/spinlock.h> | |
1da177e4 | 29 | #include <linux/net.h> |
3fc7e8a6 | 30 | #include <linux/textsearch.h> |
1da177e4 | 31 | #include <net/checksum.h> |
a80958f4 | 32 | #include <linux/rcupdate.h> |
b7aa0bf7 | 33 | #include <linux/hrtimer.h> |
131ea667 | 34 | #include <linux/dma-mapping.h> |
c8f44aff | 35 | #include <linux/netdev_features.h> |
363ec392 | 36 | #include <linux/sched.h> |
1bd758eb | 37 | #include <net/flow_dissector.h> |
a60e3cc7 | 38 | #include <linux/splice.h> |
72b31f72 | 39 | #include <linux/in6.h> |
f70ea018 | 40 | #include <net/flow.h> |
1da177e4 | 41 | |
7a6ae71b TH |
42 | /* The interface for checksum offload between the stack and networking drivers |
43 | * is as follows... | |
44 | * | |
45 | * A. IP checksum related features | |
46 | * | |
47 | * Drivers advertise checksum offload capabilities in the features of a device. | |
48 | * From the stack's point of view these are capabilities offered by the driver, | |
49 | * a driver typically only advertises features that it is capable of offloading | |
50 | * to its device. | |
51 | * | |
52 | * The checksum related features are: | |
53 | * | |
54 | * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one | |
55 | * IP (one's complement) checksum for any combination | |
56 | * of protocols or protocol layering. The checksum is | |
57 | * computed and set in a packet per the CHECKSUM_PARTIAL | |
58 | * interface (see below). | |
59 | * | |
60 | * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain | |
61 | * TCP or UDP packets over IPv4. These are specifically | |
62 | * unencapsulated packets of the form IPv4|TCP or | |
63 | * IPv4|UDP where the Protocol field in the IPv4 header | |
64 | * is TCP or UDP. The IPv4 header may contain IP options | |
65 | * This feature cannot be set in features for a device | |
66 | * with NETIF_F_HW_CSUM also set. This feature is being | |
67 | * DEPRECATED (see below). | |
68 | * | |
69 | * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain | |
70 | * TCP or UDP packets over IPv6. These are specifically | |
71 | * unencapsulated packets of the form IPv6|TCP or | |
72 | * IPv4|UDP where the Next Header field in the IPv6 | |
73 | * header is either TCP or UDP. IPv6 extension headers | |
74 | * are not supported with this feature. This feature | |
75 | * cannot be set in features for a device with | |
76 | * NETIF_F_HW_CSUM also set. This feature is being | |
77 | * DEPRECATED (see below). | |
78 | * | |
79 | * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload. | |
80 | * This flag is used only used to disable the RX checksum | |
81 | * feature for a device. The stack will accept receive | |
82 | * checksum indication in packets received on a device | |
83 | * regardless of whether NETIF_F_RXCSUM is set. | |
84 | * | |
85 | * B. Checksumming of received packets by device. Indication of checksum | |
86 | * verification is in set skb->ip_summed. Possible values are: | |
78ea85f1 DB |
87 | * |
88 | * CHECKSUM_NONE: | |
89 | * | |
7a6ae71b | 90 | * Device did not checksum this packet e.g. due to lack of capabilities. |
78ea85f1 DB |
91 | * The packet contains full (though not verified) checksum in packet but |
92 | * not in skb->csum. Thus, skb->csum is undefined in this case. | |
93 | * | |
94 | * CHECKSUM_UNNECESSARY: | |
95 | * | |
96 | * The hardware you're dealing with doesn't calculate the full checksum | |
97 | * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums | |
77cffe23 TH |
98 | * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY |
99 | * if their checksums are okay. skb->csum is still undefined in this case | |
7a6ae71b TH |
100 | * though. A driver or device must never modify the checksum field in the |
101 | * packet even if checksum is verified. | |
77cffe23 TH |
102 | * |
103 | * CHECKSUM_UNNECESSARY is applicable to following protocols: | |
104 | * TCP: IPv6 and IPv4. | |
105 | * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a | |
106 | * zero UDP checksum for either IPv4 or IPv6, the networking stack | |
107 | * may perform further validation in this case. | |
108 | * GRE: only if the checksum is present in the header. | |
109 | * SCTP: indicates the CRC in SCTP header has been validated. | |
110 | * | |
111 | * skb->csum_level indicates the number of consecutive checksums found in | |
112 | * the packet minus one that have been verified as CHECKSUM_UNNECESSARY. | |
113 | * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet | |
114 | * and a device is able to verify the checksums for UDP (possibly zero), | |
115 | * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to | |
116 | * two. If the device were only able to verify the UDP checksum and not | |
117 | * GRE, either because it doesn't support GRE checksum of because GRE | |
118 | * checksum is bad, skb->csum_level would be set to zero (TCP checksum is | |
119 | * not considered in this case). | |
78ea85f1 DB |
120 | * |
121 | * CHECKSUM_COMPLETE: | |
122 | * | |
123 | * This is the most generic way. The device supplied checksum of the _whole_ | |
124 | * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the | |
125 | * hardware doesn't need to parse L3/L4 headers to implement this. | |
126 | * | |
127 | * Note: Even if device supports only some protocols, but is able to produce | |
128 | * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY. | |
129 | * | |
130 | * CHECKSUM_PARTIAL: | |
131 | * | |
6edec0e6 TH |
132 | * A checksum is set up to be offloaded to a device as described in the |
133 | * output description for CHECKSUM_PARTIAL. This may occur on a packet | |
78ea85f1 | 134 | * received directly from another Linux OS, e.g., a virtualized Linux kernel |
6edec0e6 TH |
135 | * on the same host, or it may be set in the input path in GRO or remote |
136 | * checksum offload. For the purposes of checksum verification, the checksum | |
137 | * referred to by skb->csum_start + skb->csum_offset and any preceding | |
138 | * checksums in the packet are considered verified. Any checksums in the | |
139 | * packet that are after the checksum being offloaded are not considered to | |
140 | * be verified. | |
78ea85f1 | 141 | * |
7a6ae71b TH |
142 | * C. Checksumming on transmit for non-GSO. The stack requests checksum offload |
143 | * in the skb->ip_summed for a packet. Values are: | |
78ea85f1 DB |
144 | * |
145 | * CHECKSUM_PARTIAL: | |
146 | * | |
7a6ae71b | 147 | * The driver is required to checksum the packet as seen by hard_start_xmit() |
78ea85f1 | 148 | * from skb->csum_start up to the end, and to record/write the checksum at |
7a6ae71b TH |
149 | * offset skb->csum_start + skb->csum_offset. A driver may verify that the |
150 | * csum_start and csum_offset values are valid values given the length and | |
151 | * offset of the packet, however they should not attempt to validate that the | |
152 | * checksum refers to a legitimate transport layer checksum-- it is the | |
153 | * purview of the stack to validate that csum_start and csum_offset are set | |
154 | * correctly. | |
155 | * | |
156 | * When the stack requests checksum offload for a packet, the driver MUST | |
157 | * ensure that the checksum is set correctly. A driver can either offload the | |
158 | * checksum calculation to the device, or call skb_checksum_help (in the case | |
159 | * that the device does not support offload for a particular checksum). | |
160 | * | |
161 | * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of | |
162 | * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate | |
163 | * checksum offload capability. If a device has limited checksum capabilities | |
164 | * (for instance can only perform NETIF_F_IP_CSUM or NETIF_F_IPV6_CSUM as | |
165 | * described above) a helper function can be called to resolve | |
166 | * CHECKSUM_PARTIAL. The helper functions are skb_csum_off_chk*. The helper | |
167 | * function takes a spec argument that describes the protocol layer that is | |
168 | * supported for checksum offload and can be called for each packet. If a | |
169 | * packet does not match the specification for offload, skb_checksum_help | |
170 | * is called to resolve the checksum. | |
78ea85f1 | 171 | * |
7a6ae71b | 172 | * CHECKSUM_NONE: |
78ea85f1 | 173 | * |
7a6ae71b TH |
174 | * The skb was already checksummed by the protocol, or a checksum is not |
175 | * required. | |
78ea85f1 DB |
176 | * |
177 | * CHECKSUM_UNNECESSARY: | |
178 | * | |
7a6ae71b TH |
179 | * This has the same meaning on as CHECKSUM_NONE for checksum offload on |
180 | * output. | |
78ea85f1 | 181 | * |
7a6ae71b TH |
182 | * CHECKSUM_COMPLETE: |
183 | * Not used in checksum output. If a driver observes a packet with this value | |
184 | * set in skbuff, if should treat as CHECKSUM_NONE being set. | |
185 | * | |
186 | * D. Non-IP checksum (CRC) offloads | |
187 | * | |
188 | * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of | |
189 | * offloading the SCTP CRC in a packet. To perform this offload the stack | |
190 | * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset | |
191 | * accordingly. Note the there is no indication in the skbuff that the | |
192 | * CHECKSUM_PARTIAL refers to an SCTP checksum, a driver that supports | |
193 | * both IP checksum offload and SCTP CRC offload must verify which offload | |
194 | * is configured for a packet presumably by inspecting packet headers. | |
195 | * | |
196 | * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of | |
197 | * offloading the FCOE CRC in a packet. To perform this offload the stack | |
198 | * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset | |
199 | * accordingly. Note the there is no indication in the skbuff that the | |
200 | * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports | |
201 | * both IP checksum offload and FCOE CRC offload must verify which offload | |
202 | * is configured for a packet presumably by inspecting packet headers. | |
203 | * | |
204 | * E. Checksumming on output with GSO. | |
205 | * | |
206 | * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload | |
207 | * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the | |
208 | * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as | |
209 | * part of the GSO operation is implied. If a checksum is being offloaded | |
210 | * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset | |
211 | * are set to refer to the outermost checksum being offload (two offloaded | |
212 | * checksums are possible with UDP encapsulation). | |
78ea85f1 DB |
213 | */ |
214 | ||
60476372 | 215 | /* Don't change this without changing skb_csum_unnecessary! */ |
78ea85f1 DB |
216 | #define CHECKSUM_NONE 0 |
217 | #define CHECKSUM_UNNECESSARY 1 | |
218 | #define CHECKSUM_COMPLETE 2 | |
219 | #define CHECKSUM_PARTIAL 3 | |
1da177e4 | 220 | |
77cffe23 TH |
221 | /* Maximum value in skb->csum_level */ |
222 | #define SKB_MAX_CSUM_LEVEL 3 | |
223 | ||
0bec8c88 | 224 | #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES) |
fc910a27 | 225 | #define SKB_WITH_OVERHEAD(X) \ |
deea84b0 | 226 | ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) |
fc910a27 DM |
227 | #define SKB_MAX_ORDER(X, ORDER) \ |
228 | SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X)) | |
1da177e4 LT |
229 | #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0)) |
230 | #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2)) | |
231 | ||
87fb4b7b ED |
232 | /* return minimum truesize of one skb containing X bytes of data */ |
233 | #define SKB_TRUESIZE(X) ((X) + \ | |
234 | SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \ | |
235 | SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) | |
236 | ||
1da177e4 | 237 | struct net_device; |
716ea3a7 | 238 | struct scatterlist; |
9c55e01c | 239 | struct pipe_inode_info; |
a8f820aa | 240 | struct iov_iter; |
fd11a83d | 241 | struct napi_struct; |
1da177e4 | 242 | |
5f79e0f9 | 243 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) |
1da177e4 LT |
244 | struct nf_conntrack { |
245 | atomic_t use; | |
1da177e4 | 246 | }; |
5f79e0f9 | 247 | #endif |
1da177e4 | 248 | |
34666d46 | 249 | #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) |
1da177e4 | 250 | struct nf_bridge_info { |
bf1ac5ca | 251 | atomic_t use; |
3eaf4025 FW |
252 | enum { |
253 | BRNF_PROTO_UNCHANGED, | |
254 | BRNF_PROTO_8021Q, | |
255 | BRNF_PROTO_PPPOE | |
7fb48c5b | 256 | } orig_proto:8; |
72b1e5e4 FW |
257 | u8 pkt_otherhost:1; |
258 | u8 in_prerouting:1; | |
259 | u8 bridged_dnat:1; | |
411ffb4f | 260 | __u16 frag_max_size; |
bf1ac5ca | 261 | struct net_device *physindev; |
63cdbc06 FW |
262 | |
263 | /* always valid & non-NULL from FORWARD on, for physdev match */ | |
264 | struct net_device *physoutdev; | |
7fb48c5b | 265 | union { |
72b1e5e4 | 266 | /* prerouting: detect dnat in orig/reply direction */ |
72b31f72 BT |
267 | __be32 ipv4_daddr; |
268 | struct in6_addr ipv6_daddr; | |
72b1e5e4 FW |
269 | |
270 | /* after prerouting + nat detected: store original source | |
271 | * mac since neigh resolution overwrites it, only used while | |
272 | * skb is out in neigh layer. | |
273 | */ | |
274 | char neigh_header[8]; | |
72b31f72 | 275 | }; |
1da177e4 LT |
276 | }; |
277 | #endif | |
278 | ||
1da177e4 LT |
279 | struct sk_buff_head { |
280 | /* These two members must be first. */ | |
281 | struct sk_buff *next; | |
282 | struct sk_buff *prev; | |
283 | ||
284 | __u32 qlen; | |
285 | spinlock_t lock; | |
286 | }; | |
287 | ||
288 | struct sk_buff; | |
289 | ||
9d4dde52 IC |
290 | /* To allow 64K frame to be packed as single skb without frag_list we |
291 | * require 64K/PAGE_SIZE pages plus 1 additional page to allow for | |
292 | * buffers which do not start on a page boundary. | |
293 | * | |
294 | * Since GRO uses frags we allocate at least 16 regardless of page | |
295 | * size. | |
a715dea3 | 296 | */ |
9d4dde52 | 297 | #if (65536/PAGE_SIZE + 1) < 16 |
eec00954 | 298 | #define MAX_SKB_FRAGS 16UL |
a715dea3 | 299 | #else |
9d4dde52 | 300 | #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1) |
a715dea3 | 301 | #endif |
1da177e4 LT |
302 | |
303 | typedef struct skb_frag_struct skb_frag_t; | |
304 | ||
305 | struct skb_frag_struct { | |
a8605c60 IC |
306 | struct { |
307 | struct page *p; | |
308 | } page; | |
cb4dfe56 | 309 | #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536) |
a309bb07 DM |
310 | __u32 page_offset; |
311 | __u32 size; | |
cb4dfe56 ED |
312 | #else |
313 | __u16 page_offset; | |
314 | __u16 size; | |
315 | #endif | |
1da177e4 LT |
316 | }; |
317 | ||
9e903e08 ED |
318 | static inline unsigned int skb_frag_size(const skb_frag_t *frag) |
319 | { | |
320 | return frag->size; | |
321 | } | |
322 | ||
323 | static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size) | |
324 | { | |
325 | frag->size = size; | |
326 | } | |
327 | ||
328 | static inline void skb_frag_size_add(skb_frag_t *frag, int delta) | |
329 | { | |
330 | frag->size += delta; | |
331 | } | |
332 | ||
333 | static inline void skb_frag_size_sub(skb_frag_t *frag, int delta) | |
334 | { | |
335 | frag->size -= delta; | |
336 | } | |
337 | ||
ac45f602 PO |
338 | #define HAVE_HW_TIME_STAMP |
339 | ||
340 | /** | |
d3a21be8 | 341 | * struct skb_shared_hwtstamps - hardware time stamps |
ac45f602 PO |
342 | * @hwtstamp: hardware time stamp transformed into duration |
343 | * since arbitrary point in time | |
ac45f602 PO |
344 | * |
345 | * Software time stamps generated by ktime_get_real() are stored in | |
4d276eb6 | 346 | * skb->tstamp. |
ac45f602 PO |
347 | * |
348 | * hwtstamps can only be compared against other hwtstamps from | |
349 | * the same device. | |
350 | * | |
351 | * This structure is attached to packets as part of the | |
352 | * &skb_shared_info. Use skb_hwtstamps() to get a pointer. | |
353 | */ | |
354 | struct skb_shared_hwtstamps { | |
355 | ktime_t hwtstamp; | |
ac45f602 PO |
356 | }; |
357 | ||
2244d07b OH |
358 | /* Definitions for tx_flags in struct skb_shared_info */ |
359 | enum { | |
360 | /* generate hardware time stamp */ | |
361 | SKBTX_HW_TSTAMP = 1 << 0, | |
362 | ||
e7fd2885 | 363 | /* generate software time stamp when queueing packet to NIC */ |
2244d07b OH |
364 | SKBTX_SW_TSTAMP = 1 << 1, |
365 | ||
366 | /* device driver is going to provide hardware time stamp */ | |
367 | SKBTX_IN_PROGRESS = 1 << 2, | |
368 | ||
a6686f2f | 369 | /* device driver supports TX zero-copy buffers */ |
62b1a8ab | 370 | SKBTX_DEV_ZEROCOPY = 1 << 3, |
6e3e939f JB |
371 | |
372 | /* generate wifi status information (where possible) */ | |
62b1a8ab | 373 | SKBTX_WIFI_STATUS = 1 << 4, |
c9af6db4 PS |
374 | |
375 | /* This indicates at least one fragment might be overwritten | |
376 | * (as in vmsplice(), sendfile() ...) | |
377 | * If we need to compute a TX checksum, we'll need to copy | |
378 | * all frags to avoid possible bad checksum | |
379 | */ | |
380 | SKBTX_SHARED_FRAG = 1 << 5, | |
e7fd2885 WB |
381 | |
382 | /* generate software time stamp when entering packet scheduling */ | |
383 | SKBTX_SCHED_TSTAMP = 1 << 6, | |
e1c8a607 WB |
384 | |
385 | /* generate software timestamp on peer data acknowledgment */ | |
386 | SKBTX_ACK_TSTAMP = 1 << 7, | |
a6686f2f SM |
387 | }; |
388 | ||
e1c8a607 WB |
389 | #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \ |
390 | SKBTX_SCHED_TSTAMP | \ | |
391 | SKBTX_ACK_TSTAMP) | |
f24b9be5 WB |
392 | #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP) |
393 | ||
a6686f2f SM |
394 | /* |
395 | * The callback notifies userspace to release buffers when skb DMA is done in | |
396 | * lower device, the skb last reference should be 0 when calling this. | |
e19d6763 MT |
397 | * The zerocopy_success argument is true if zero copy transmit occurred, |
398 | * false on data copy or out of memory error caused by data copy attempt. | |
ca8f4fb2 MT |
399 | * The ctx field is used to track device context. |
400 | * The desc field is used to track userspace buffer index. | |
a6686f2f SM |
401 | */ |
402 | struct ubuf_info { | |
e19d6763 | 403 | void (*callback)(struct ubuf_info *, bool zerocopy_success); |
ca8f4fb2 | 404 | void *ctx; |
a6686f2f | 405 | unsigned long desc; |
ac45f602 PO |
406 | }; |
407 | ||
1da177e4 LT |
408 | /* This data is invariant across clones and lives at |
409 | * the end of the header data, ie. at skb->end. | |
410 | */ | |
411 | struct skb_shared_info { | |
9f42f126 IC |
412 | unsigned char nr_frags; |
413 | __u8 tx_flags; | |
7967168c HX |
414 | unsigned short gso_size; |
415 | /* Warning: this field is not always filled in (UFO)! */ | |
416 | unsigned short gso_segs; | |
417 | unsigned short gso_type; | |
1da177e4 | 418 | struct sk_buff *frag_list; |
ac45f602 | 419 | struct skb_shared_hwtstamps hwtstamps; |
09c2d251 | 420 | u32 tskey; |
9f42f126 | 421 | __be32 ip6_frag_id; |
ec7d2f2c ED |
422 | |
423 | /* | |
424 | * Warning : all fields before dataref are cleared in __alloc_skb() | |
425 | */ | |
426 | atomic_t dataref; | |
427 | ||
69e3c75f JB |
428 | /* Intermediate layers must ensure that destructor_arg |
429 | * remains valid until skb destructor */ | |
430 | void * destructor_arg; | |
a6686f2f | 431 | |
fed66381 ED |
432 | /* must be last field, see pskb_expand_head() */ |
433 | skb_frag_t frags[MAX_SKB_FRAGS]; | |
1da177e4 LT |
434 | }; |
435 | ||
436 | /* We divide dataref into two halves. The higher 16 bits hold references | |
437 | * to the payload part of skb->data. The lower 16 bits hold references to | |
334a8132 PM |
438 | * the entire skb->data. A clone of a headerless skb holds the length of |
439 | * the header in skb->hdr_len. | |
1da177e4 LT |
440 | * |
441 | * All users must obey the rule that the skb->data reference count must be | |
442 | * greater than or equal to the payload reference count. | |
443 | * | |
444 | * Holding a reference to the payload part means that the user does not | |
445 | * care about modifications to the header part of skb->data. | |
446 | */ | |
447 | #define SKB_DATAREF_SHIFT 16 | |
448 | #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1) | |
449 | ||
d179cd12 DM |
450 | |
451 | enum { | |
c8753d55 VS |
452 | SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */ |
453 | SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */ | |
454 | SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */ | |
d179cd12 DM |
455 | }; |
456 | ||
7967168c HX |
457 | enum { |
458 | SKB_GSO_TCPV4 = 1 << 0, | |
f83ef8c0 | 459 | SKB_GSO_UDP = 1 << 1, |
576a30eb HX |
460 | |
461 | /* This indicates the skb is from an untrusted source. */ | |
462 | SKB_GSO_DODGY = 1 << 2, | |
b0da8537 MC |
463 | |
464 | /* This indicates the tcp segment has CWR set. */ | |
f83ef8c0 HX |
465 | SKB_GSO_TCP_ECN = 1 << 3, |
466 | ||
467 | SKB_GSO_TCPV6 = 1 << 4, | |
01d5b2fc CL |
468 | |
469 | SKB_GSO_FCOE = 1 << 5, | |
68c33163 PS |
470 | |
471 | SKB_GSO_GRE = 1 << 6, | |
73136267 | 472 | |
4b28252c | 473 | SKB_GSO_GRE_CSUM = 1 << 7, |
0d89d203 | 474 | |
4b28252c | 475 | SKB_GSO_IPIP = 1 << 8, |
cb32f511 | 476 | |
4b28252c | 477 | SKB_GSO_SIT = 1 << 9, |
61c1db7f | 478 | |
4b28252c | 479 | SKB_GSO_UDP_TUNNEL = 1 << 10, |
0f4f4ffa TH |
480 | |
481 | SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11, | |
4749c09c | 482 | |
59b93b41 | 483 | SKB_GSO_TUNNEL_REMCSUM = 1 << 12, |
7967168c HX |
484 | }; |
485 | ||
2e07fa9c ACM |
486 | #if BITS_PER_LONG > 32 |
487 | #define NET_SKBUFF_DATA_USES_OFFSET 1 | |
488 | #endif | |
489 | ||
490 | #ifdef NET_SKBUFF_DATA_USES_OFFSET | |
491 | typedef unsigned int sk_buff_data_t; | |
492 | #else | |
493 | typedef unsigned char *sk_buff_data_t; | |
494 | #endif | |
495 | ||
363ec392 ED |
496 | /** |
497 | * struct skb_mstamp - multi resolution time stamps | |
498 | * @stamp_us: timestamp in us resolution | |
499 | * @stamp_jiffies: timestamp in jiffies | |
500 | */ | |
501 | struct skb_mstamp { | |
502 | union { | |
503 | u64 v64; | |
504 | struct { | |
505 | u32 stamp_us; | |
506 | u32 stamp_jiffies; | |
507 | }; | |
508 | }; | |
509 | }; | |
510 | ||
511 | /** | |
512 | * skb_mstamp_get - get current timestamp | |
513 | * @cl: place to store timestamps | |
514 | */ | |
515 | static inline void skb_mstamp_get(struct skb_mstamp *cl) | |
516 | { | |
517 | u64 val = local_clock(); | |
518 | ||
519 | do_div(val, NSEC_PER_USEC); | |
520 | cl->stamp_us = (u32)val; | |
521 | cl->stamp_jiffies = (u32)jiffies; | |
522 | } | |
523 | ||
524 | /** | |
525 | * skb_mstamp_delta - compute the difference in usec between two skb_mstamp | |
526 | * @t1: pointer to newest sample | |
527 | * @t0: pointer to oldest sample | |
528 | */ | |
529 | static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1, | |
530 | const struct skb_mstamp *t0) | |
531 | { | |
532 | s32 delta_us = t1->stamp_us - t0->stamp_us; | |
533 | u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies; | |
534 | ||
535 | /* If delta_us is negative, this might be because interval is too big, | |
536 | * or local_clock() drift is too big : fallback using jiffies. | |
537 | */ | |
538 | if (delta_us <= 0 || | |
539 | delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ))) | |
540 | ||
541 | delta_us = jiffies_to_usecs(delta_jiffies); | |
542 | ||
543 | return delta_us; | |
544 | } | |
545 | ||
625a5e10 YC |
546 | static inline bool skb_mstamp_after(const struct skb_mstamp *t1, |
547 | const struct skb_mstamp *t0) | |
548 | { | |
549 | s32 diff = t1->stamp_jiffies - t0->stamp_jiffies; | |
550 | ||
551 | if (!diff) | |
552 | diff = t1->stamp_us - t0->stamp_us; | |
553 | return diff > 0; | |
554 | } | |
363ec392 | 555 | |
1da177e4 LT |
556 | /** |
557 | * struct sk_buff - socket buffer | |
558 | * @next: Next buffer in list | |
559 | * @prev: Previous buffer in list | |
363ec392 | 560 | * @tstamp: Time we arrived/left |
56b17425 | 561 | * @rbnode: RB tree node, alternative to next/prev for netem/tcp |
d84e0bd7 | 562 | * @sk: Socket we are owned by |
1da177e4 | 563 | * @dev: Device we arrived on/are leaving by |
d84e0bd7 | 564 | * @cb: Control buffer. Free for use by every layer. Put private vars here |
7fee226a | 565 | * @_skb_refdst: destination entry (with norefcount bit) |
67be2dd1 | 566 | * @sp: the security path, used for xfrm |
1da177e4 LT |
567 | * @len: Length of actual data |
568 | * @data_len: Data length | |
569 | * @mac_len: Length of link layer header | |
334a8132 | 570 | * @hdr_len: writable header length of cloned skb |
663ead3b HX |
571 | * @csum: Checksum (must include start/offset pair) |
572 | * @csum_start: Offset from skb->head where checksumming should start | |
573 | * @csum_offset: Offset from csum_start where checksum should be stored | |
d84e0bd7 | 574 | * @priority: Packet queueing priority |
60ff7467 | 575 | * @ignore_df: allow local fragmentation |
1da177e4 | 576 | * @cloned: Head may be cloned (check refcnt to be sure) |
d84e0bd7 | 577 | * @ip_summed: Driver fed us an IP checksum |
1da177e4 | 578 | * @nohdr: Payload reference only, must not modify header |
d84e0bd7 | 579 | * @nfctinfo: Relationship of this skb to the connection |
1da177e4 | 580 | * @pkt_type: Packet class |
c83c2486 | 581 | * @fclone: skbuff clone status |
c83c2486 | 582 | * @ipvs_property: skbuff is owned by ipvs |
31729363 RD |
583 | * @peeked: this packet has been seen already, so stats have been |
584 | * done for it, don't do them again | |
ba9dda3a | 585 | * @nf_trace: netfilter packet trace flag |
d84e0bd7 DB |
586 | * @protocol: Packet protocol from driver |
587 | * @destructor: Destruct function | |
588 | * @nfct: Associated connection, if any | |
1da177e4 | 589 | * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c |
8964be4a | 590 | * @skb_iif: ifindex of device we arrived on |
1da177e4 LT |
591 | * @tc_index: Traffic control index |
592 | * @tc_verd: traffic control verdict | |
61b905da | 593 | * @hash: the packet hash |
d84e0bd7 | 594 | * @queue_mapping: Queue mapping for multiqueue devices |
0b725a2c | 595 | * @xmit_more: More SKBs are pending for this queue |
553a5672 | 596 | * @ndisc_nodetype: router type (from link layer) |
d84e0bd7 | 597 | * @ooo_okay: allow the mapping of a socket to a queue to be changed |
61b905da | 598 | * @l4_hash: indicate hash is a canonical 4-tuple hash over transport |
4ca2462e | 599 | * ports. |
a3b18ddb | 600 | * @sw_hash: indicates hash was computed in software stack |
6e3e939f JB |
601 | * @wifi_acked_valid: wifi_acked was set |
602 | * @wifi_acked: whether frame was acked on wifi or not | |
3bdc0eba | 603 | * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS |
06021292 | 604 | * @napi_id: id of the NAPI struct this skb came from |
984bc16c | 605 | * @secmark: security marking |
0c4f691f | 606 | * @offload_fwd_mark: fwding offload mark |
d84e0bd7 | 607 | * @mark: Generic packet mark |
86a9bad3 | 608 | * @vlan_proto: vlan encapsulation protocol |
6aa895b0 | 609 | * @vlan_tci: vlan tag control information |
0d89d203 | 610 | * @inner_protocol: Protocol (encapsulation) |
6a674e9c JG |
611 | * @inner_transport_header: Inner transport layer header (encapsulation) |
612 | * @inner_network_header: Network layer header (encapsulation) | |
aefbd2b3 | 613 | * @inner_mac_header: Link layer header (encapsulation) |
d84e0bd7 DB |
614 | * @transport_header: Transport layer header |
615 | * @network_header: Network layer header | |
616 | * @mac_header: Link layer header | |
617 | * @tail: Tail pointer | |
618 | * @end: End pointer | |
619 | * @head: Head of buffer | |
620 | * @data: Data head pointer | |
621 | * @truesize: Buffer size | |
622 | * @users: User count - see {datagram,tcp}.c | |
1da177e4 LT |
623 | */ |
624 | ||
625 | struct sk_buff { | |
363ec392 | 626 | union { |
56b17425 ED |
627 | struct { |
628 | /* These two members must be first. */ | |
629 | struct sk_buff *next; | |
630 | struct sk_buff *prev; | |
631 | ||
632 | union { | |
633 | ktime_t tstamp; | |
634 | struct skb_mstamp skb_mstamp; | |
635 | }; | |
636 | }; | |
637 | struct rb_node rbnode; /* used in netem & tcp stack */ | |
363ec392 | 638 | }; |
da3f5cf1 | 639 | struct sock *sk; |
1da177e4 | 640 | struct net_device *dev; |
1da177e4 | 641 | |
1da177e4 LT |
642 | /* |
643 | * This is the control buffer. It is free to use for every | |
644 | * layer. Please put your private variables there. If you | |
645 | * want to keep them across layers you have to do a skb_clone() | |
646 | * first. This is owned by whoever has the skb queued ATM. | |
647 | */ | |
da3f5cf1 | 648 | char cb[48] __aligned(8); |
1da177e4 | 649 | |
7fee226a | 650 | unsigned long _skb_refdst; |
b1937227 | 651 | void (*destructor)(struct sk_buff *skb); |
da3f5cf1 FF |
652 | #ifdef CONFIG_XFRM |
653 | struct sec_path *sp; | |
b1937227 ED |
654 | #endif |
655 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) | |
656 | struct nf_conntrack *nfct; | |
657 | #endif | |
85224844 | 658 | #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) |
b1937227 | 659 | struct nf_bridge_info *nf_bridge; |
da3f5cf1 | 660 | #endif |
1da177e4 | 661 | unsigned int len, |
334a8132 PM |
662 | data_len; |
663 | __u16 mac_len, | |
664 | hdr_len; | |
b1937227 ED |
665 | |
666 | /* Following fields are _not_ copied in __copy_skb_header() | |
667 | * Note that queue_mapping is here mostly to fill a hole. | |
668 | */ | |
fe55f6d5 | 669 | kmemcheck_bitfield_begin(flags1); |
b1937227 ED |
670 | __u16 queue_mapping; |
671 | __u8 cloned:1, | |
6869c4d8 | 672 | nohdr:1, |
b84f4cc9 | 673 | fclone:2, |
a59322be | 674 | peeked:1, |
b1937227 ED |
675 | head_frag:1, |
676 | xmit_more:1; | |
677 | /* one bit hole */ | |
fe55f6d5 | 678 | kmemcheck_bitfield_end(flags1); |
4031ae6e | 679 | |
b1937227 ED |
680 | /* fields enclosed in headers_start/headers_end are copied |
681 | * using a single memcpy() in __copy_skb_header() | |
682 | */ | |
ebcf34f3 | 683 | /* private: */ |
b1937227 | 684 | __u32 headers_start[0]; |
ebcf34f3 | 685 | /* public: */ |
4031ae6e | 686 | |
233577a2 HFS |
687 | /* if you move pkt_type around you also must adapt those constants */ |
688 | #ifdef __BIG_ENDIAN_BITFIELD | |
689 | #define PKT_TYPE_MAX (7 << 5) | |
690 | #else | |
691 | #define PKT_TYPE_MAX 7 | |
1da177e4 | 692 | #endif |
233577a2 | 693 | #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset) |
fe55f6d5 | 694 | |
233577a2 | 695 | __u8 __pkt_type_offset[0]; |
b1937227 | 696 | __u8 pkt_type:3; |
c93bdd0e | 697 | __u8 pfmemalloc:1; |
b1937227 ED |
698 | __u8 ignore_df:1; |
699 | __u8 nfctinfo:3; | |
700 | ||
701 | __u8 nf_trace:1; | |
702 | __u8 ip_summed:2; | |
3853b584 | 703 | __u8 ooo_okay:1; |
61b905da | 704 | __u8 l4_hash:1; |
a3b18ddb | 705 | __u8 sw_hash:1; |
6e3e939f JB |
706 | __u8 wifi_acked_valid:1; |
707 | __u8 wifi_acked:1; | |
b1937227 | 708 | |
3bdc0eba | 709 | __u8 no_fcs:1; |
77cffe23 | 710 | /* Indicates the inner headers are valid in the skbuff. */ |
6a674e9c | 711 | __u8 encapsulation:1; |
7e2b10c1 | 712 | __u8 encap_hdr_csum:1; |
5d0c2b95 | 713 | __u8 csum_valid:1; |
7e3cead5 | 714 | __u8 csum_complete_sw:1; |
b1937227 ED |
715 | __u8 csum_level:2; |
716 | __u8 csum_bad:1; | |
fe55f6d5 | 717 | |
b1937227 ED |
718 | #ifdef CONFIG_IPV6_NDISC_NODETYPE |
719 | __u8 ndisc_nodetype:2; | |
720 | #endif | |
721 | __u8 ipvs_property:1; | |
8bce6d7d | 722 | __u8 inner_protocol_type:1; |
e585f236 TH |
723 | __u8 remcsum_offload:1; |
724 | /* 3 or 5 bit hole */ | |
b1937227 ED |
725 | |
726 | #ifdef CONFIG_NET_SCHED | |
727 | __u16 tc_index; /* traffic control index */ | |
728 | #ifdef CONFIG_NET_CLS_ACT | |
729 | __u16 tc_verd; /* traffic control verdict */ | |
730 | #endif | |
731 | #endif | |
fe55f6d5 | 732 | |
b1937227 ED |
733 | union { |
734 | __wsum csum; | |
735 | struct { | |
736 | __u16 csum_start; | |
737 | __u16 csum_offset; | |
738 | }; | |
739 | }; | |
740 | __u32 priority; | |
741 | int skb_iif; | |
742 | __u32 hash; | |
743 | __be16 vlan_proto; | |
744 | __u16 vlan_tci; | |
2bd82484 ED |
745 | #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS) |
746 | union { | |
747 | unsigned int napi_id; | |
748 | unsigned int sender_cpu; | |
749 | }; | |
97fc2f08 | 750 | #endif |
0c4f691f | 751 | union { |
984bc16c | 752 | #ifdef CONFIG_NETWORK_SECMARK |
0c4f691f SF |
753 | __u32 secmark; |
754 | #endif | |
755 | #ifdef CONFIG_NET_SWITCHDEV | |
756 | __u32 offload_fwd_mark; | |
984bc16c | 757 | #endif |
0c4f691f SF |
758 | }; |
759 | ||
3b885787 NH |
760 | union { |
761 | __u32 mark; | |
16fad69c | 762 | __u32 reserved_tailroom; |
3b885787 | 763 | }; |
1da177e4 | 764 | |
8bce6d7d TH |
765 | union { |
766 | __be16 inner_protocol; | |
767 | __u8 inner_ipproto; | |
768 | }; | |
769 | ||
1a37e412 SH |
770 | __u16 inner_transport_header; |
771 | __u16 inner_network_header; | |
772 | __u16 inner_mac_header; | |
b1937227 ED |
773 | |
774 | __be16 protocol; | |
1a37e412 SH |
775 | __u16 transport_header; |
776 | __u16 network_header; | |
777 | __u16 mac_header; | |
b1937227 | 778 | |
ebcf34f3 | 779 | /* private: */ |
b1937227 | 780 | __u32 headers_end[0]; |
ebcf34f3 | 781 | /* public: */ |
b1937227 | 782 | |
1da177e4 | 783 | /* These elements must be at the end, see alloc_skb() for details. */ |
27a884dc | 784 | sk_buff_data_t tail; |
4305b541 | 785 | sk_buff_data_t end; |
1da177e4 | 786 | unsigned char *head, |
4305b541 | 787 | *data; |
27a884dc ACM |
788 | unsigned int truesize; |
789 | atomic_t users; | |
1da177e4 LT |
790 | }; |
791 | ||
792 | #ifdef __KERNEL__ | |
793 | /* | |
794 | * Handling routines are only of interest to the kernel | |
795 | */ | |
796 | #include <linux/slab.h> | |
797 | ||
1da177e4 | 798 | |
c93bdd0e MG |
799 | #define SKB_ALLOC_FCLONE 0x01 |
800 | #define SKB_ALLOC_RX 0x02 | |
fd11a83d | 801 | #define SKB_ALLOC_NAPI 0x04 |
c93bdd0e MG |
802 | |
803 | /* Returns true if the skb was allocated from PFMEMALLOC reserves */ | |
804 | static inline bool skb_pfmemalloc(const struct sk_buff *skb) | |
805 | { | |
806 | return unlikely(skb->pfmemalloc); | |
807 | } | |
808 | ||
7fee226a ED |
809 | /* |
810 | * skb might have a dst pointer attached, refcounted or not. | |
811 | * _skb_refdst low order bit is set if refcount was _not_ taken | |
812 | */ | |
813 | #define SKB_DST_NOREF 1UL | |
814 | #define SKB_DST_PTRMASK ~(SKB_DST_NOREF) | |
815 | ||
816 | /** | |
817 | * skb_dst - returns skb dst_entry | |
818 | * @skb: buffer | |
819 | * | |
820 | * Returns skb dst_entry, regardless of reference taken or not. | |
821 | */ | |
adf30907 ED |
822 | static inline struct dst_entry *skb_dst(const struct sk_buff *skb) |
823 | { | |
7fee226a ED |
824 | /* If refdst was not refcounted, check we still are in a |
825 | * rcu_read_lock section | |
826 | */ | |
827 | WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) && | |
828 | !rcu_read_lock_held() && | |
829 | !rcu_read_lock_bh_held()); | |
830 | return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK); | |
adf30907 ED |
831 | } |
832 | ||
7fee226a ED |
833 | /** |
834 | * skb_dst_set - sets skb dst | |
835 | * @skb: buffer | |
836 | * @dst: dst entry | |
837 | * | |
838 | * Sets skb dst, assuming a reference was taken on dst and should | |
839 | * be released by skb_dst_drop() | |
840 | */ | |
adf30907 ED |
841 | static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst) |
842 | { | |
7fee226a ED |
843 | skb->_skb_refdst = (unsigned long)dst; |
844 | } | |
845 | ||
932bc4d7 JA |
846 | /** |
847 | * skb_dst_set_noref - sets skb dst, hopefully, without taking reference | |
848 | * @skb: buffer | |
849 | * @dst: dst entry | |
850 | * | |
851 | * Sets skb dst, assuming a reference was not taken on dst. | |
852 | * If dst entry is cached, we do not take reference and dst_release | |
853 | * will be avoided by refdst_drop. If dst entry is not cached, we take | |
854 | * reference, so that last dst_release can destroy the dst immediately. | |
855 | */ | |
856 | static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst) | |
857 | { | |
dbfc4fb7 HFS |
858 | WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held()); |
859 | skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF; | |
932bc4d7 | 860 | } |
7fee226a ED |
861 | |
862 | /** | |
25985edc | 863 | * skb_dst_is_noref - Test if skb dst isn't refcounted |
7fee226a ED |
864 | * @skb: buffer |
865 | */ | |
866 | static inline bool skb_dst_is_noref(const struct sk_buff *skb) | |
867 | { | |
868 | return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb); | |
adf30907 ED |
869 | } |
870 | ||
511c3f92 ED |
871 | static inline struct rtable *skb_rtable(const struct sk_buff *skb) |
872 | { | |
adf30907 | 873 | return (struct rtable *)skb_dst(skb); |
511c3f92 ED |
874 | } |
875 | ||
7965bd4d JP |
876 | void kfree_skb(struct sk_buff *skb); |
877 | void kfree_skb_list(struct sk_buff *segs); | |
878 | void skb_tx_error(struct sk_buff *skb); | |
879 | void consume_skb(struct sk_buff *skb); | |
880 | void __kfree_skb(struct sk_buff *skb); | |
d7e8883c | 881 | extern struct kmem_cache *skbuff_head_cache; |
bad43ca8 | 882 | |
7965bd4d JP |
883 | void kfree_skb_partial(struct sk_buff *skb, bool head_stolen); |
884 | bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from, | |
885 | bool *fragstolen, int *delta_truesize); | |
bad43ca8 | 886 | |
7965bd4d JP |
887 | struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags, |
888 | int node); | |
2ea2f62c | 889 | struct sk_buff *__build_skb(void *data, unsigned int frag_size); |
7965bd4d | 890 | struct sk_buff *build_skb(void *data, unsigned int frag_size); |
d179cd12 | 891 | static inline struct sk_buff *alloc_skb(unsigned int size, |
dd0fc66f | 892 | gfp_t priority) |
d179cd12 | 893 | { |
564824b0 | 894 | return __alloc_skb(size, priority, 0, NUMA_NO_NODE); |
d179cd12 DM |
895 | } |
896 | ||
2e4e4410 ED |
897 | struct sk_buff *alloc_skb_with_frags(unsigned long header_len, |
898 | unsigned long data_len, | |
899 | int max_page_order, | |
900 | int *errcode, | |
901 | gfp_t gfp_mask); | |
902 | ||
d0bf4a9e ED |
903 | /* Layout of fast clones : [skb1][skb2][fclone_ref] */ |
904 | struct sk_buff_fclones { | |
905 | struct sk_buff skb1; | |
906 | ||
907 | struct sk_buff skb2; | |
908 | ||
909 | atomic_t fclone_ref; | |
910 | }; | |
911 | ||
912 | /** | |
913 | * skb_fclone_busy - check if fclone is busy | |
914 | * @skb: buffer | |
915 | * | |
bda13fed | 916 | * Returns true if skb is a fast clone, and its clone is not freed. |
39bb5e62 ED |
917 | * Some drivers call skb_orphan() in their ndo_start_xmit(), |
918 | * so we also check that this didnt happen. | |
d0bf4a9e | 919 | */ |
39bb5e62 ED |
920 | static inline bool skb_fclone_busy(const struct sock *sk, |
921 | const struct sk_buff *skb) | |
d0bf4a9e ED |
922 | { |
923 | const struct sk_buff_fclones *fclones; | |
924 | ||
925 | fclones = container_of(skb, struct sk_buff_fclones, skb1); | |
926 | ||
927 | return skb->fclone == SKB_FCLONE_ORIG && | |
6ffe75eb | 928 | atomic_read(&fclones->fclone_ref) > 1 && |
39bb5e62 | 929 | fclones->skb2.sk == sk; |
d0bf4a9e ED |
930 | } |
931 | ||
d179cd12 | 932 | static inline struct sk_buff *alloc_skb_fclone(unsigned int size, |
dd0fc66f | 933 | gfp_t priority) |
d179cd12 | 934 | { |
c93bdd0e | 935 | return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE); |
d179cd12 DM |
936 | } |
937 | ||
7965bd4d | 938 | struct sk_buff *__alloc_skb_head(gfp_t priority, int node); |
0ebd0ac5 PM |
939 | static inline struct sk_buff *alloc_skb_head(gfp_t priority) |
940 | { | |
941 | return __alloc_skb_head(priority, -1); | |
942 | } | |
943 | ||
7965bd4d JP |
944 | struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src); |
945 | int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask); | |
946 | struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority); | |
947 | struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority); | |
bad93e9d OP |
948 | struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom, |
949 | gfp_t gfp_mask, bool fclone); | |
950 | static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom, | |
951 | gfp_t gfp_mask) | |
952 | { | |
953 | return __pskb_copy_fclone(skb, headroom, gfp_mask, false); | |
954 | } | |
7965bd4d JP |
955 | |
956 | int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask); | |
957 | struct sk_buff *skb_realloc_headroom(struct sk_buff *skb, | |
958 | unsigned int headroom); | |
959 | struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom, | |
960 | int newtailroom, gfp_t priority); | |
25a91d8d FD |
961 | int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg, |
962 | int offset, int len); | |
7965bd4d JP |
963 | int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset, |
964 | int len); | |
965 | int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer); | |
966 | int skb_pad(struct sk_buff *skb, int pad); | |
ead2ceb0 | 967 | #define dev_kfree_skb(a) consume_skb(a) |
1da177e4 | 968 | |
7965bd4d JP |
969 | int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb, |
970 | int getfrag(void *from, char *to, int offset, | |
971 | int len, int odd, struct sk_buff *skb), | |
972 | void *from, int length); | |
e89e9cf5 | 973 | |
be12a1fe HFS |
974 | int skb_append_pagefrags(struct sk_buff *skb, struct page *page, |
975 | int offset, size_t size); | |
976 | ||
d94d9fee | 977 | struct skb_seq_state { |
677e90ed TG |
978 | __u32 lower_offset; |
979 | __u32 upper_offset; | |
980 | __u32 frag_idx; | |
981 | __u32 stepped_offset; | |
982 | struct sk_buff *root_skb; | |
983 | struct sk_buff *cur_skb; | |
984 | __u8 *frag_data; | |
985 | }; | |
986 | ||
7965bd4d JP |
987 | void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from, |
988 | unsigned int to, struct skb_seq_state *st); | |
989 | unsigned int skb_seq_read(unsigned int consumed, const u8 **data, | |
990 | struct skb_seq_state *st); | |
991 | void skb_abort_seq_read(struct skb_seq_state *st); | |
677e90ed | 992 | |
7965bd4d | 993 | unsigned int skb_find_text(struct sk_buff *skb, unsigned int from, |
059a2440 | 994 | unsigned int to, struct ts_config *config); |
3fc7e8a6 | 995 | |
09323cc4 TH |
996 | /* |
997 | * Packet hash types specify the type of hash in skb_set_hash. | |
998 | * | |
999 | * Hash types refer to the protocol layer addresses which are used to | |
1000 | * construct a packet's hash. The hashes are used to differentiate or identify | |
1001 | * flows of the protocol layer for the hash type. Hash types are either | |
1002 | * layer-2 (L2), layer-3 (L3), or layer-4 (L4). | |
1003 | * | |
1004 | * Properties of hashes: | |
1005 | * | |
1006 | * 1) Two packets in different flows have different hash values | |
1007 | * 2) Two packets in the same flow should have the same hash value | |
1008 | * | |
1009 | * A hash at a higher layer is considered to be more specific. A driver should | |
1010 | * set the most specific hash possible. | |
1011 | * | |
1012 | * A driver cannot indicate a more specific hash than the layer at which a hash | |
1013 | * was computed. For instance an L3 hash cannot be set as an L4 hash. | |
1014 | * | |
1015 | * A driver may indicate a hash level which is less specific than the | |
1016 | * actual layer the hash was computed on. For instance, a hash computed | |
1017 | * at L4 may be considered an L3 hash. This should only be done if the | |
1018 | * driver can't unambiguously determine that the HW computed the hash at | |
1019 | * the higher layer. Note that the "should" in the second property above | |
1020 | * permits this. | |
1021 | */ | |
1022 | enum pkt_hash_types { | |
1023 | PKT_HASH_TYPE_NONE, /* Undefined type */ | |
1024 | PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */ | |
1025 | PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */ | |
1026 | PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */ | |
1027 | }; | |
1028 | ||
bcc83839 | 1029 | static inline void skb_clear_hash(struct sk_buff *skb) |
09323cc4 | 1030 | { |
bcc83839 | 1031 | skb->hash = 0; |
a3b18ddb | 1032 | skb->sw_hash = 0; |
bcc83839 TH |
1033 | skb->l4_hash = 0; |
1034 | } | |
1035 | ||
1036 | static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb) | |
1037 | { | |
1038 | if (!skb->l4_hash) | |
1039 | skb_clear_hash(skb); | |
1040 | } | |
1041 | ||
1042 | static inline void | |
1043 | __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4) | |
1044 | { | |
1045 | skb->l4_hash = is_l4; | |
1046 | skb->sw_hash = is_sw; | |
61b905da | 1047 | skb->hash = hash; |
09323cc4 TH |
1048 | } |
1049 | ||
bcc83839 TH |
1050 | static inline void |
1051 | skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type) | |
1052 | { | |
1053 | /* Used by drivers to set hash from HW */ | |
1054 | __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4); | |
1055 | } | |
1056 | ||
1057 | static inline void | |
1058 | __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4) | |
1059 | { | |
1060 | __skb_set_hash(skb, hash, true, is_l4); | |
1061 | } | |
1062 | ||
e5276937 TH |
1063 | void __skb_get_hash(struct sk_buff *skb); |
1064 | u32 skb_get_poff(const struct sk_buff *skb); | |
1065 | u32 __skb_get_poff(const struct sk_buff *skb, void *data, | |
1066 | const struct flow_keys *keys, int hlen); | |
1067 | __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto, | |
1068 | void *data, int hlen_proto); | |
1069 | ||
1070 | static inline __be32 skb_flow_get_ports(const struct sk_buff *skb, | |
1071 | int thoff, u8 ip_proto) | |
1072 | { | |
1073 | return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0); | |
1074 | } | |
1075 | ||
1076 | void skb_flow_dissector_init(struct flow_dissector *flow_dissector, | |
1077 | const struct flow_dissector_key *key, | |
1078 | unsigned int key_count); | |
1079 | ||
1080 | bool __skb_flow_dissect(const struct sk_buff *skb, | |
1081 | struct flow_dissector *flow_dissector, | |
1082 | void *target_container, | |
cd79a238 TH |
1083 | void *data, __be16 proto, int nhoff, int hlen, |
1084 | unsigned int flags); | |
e5276937 TH |
1085 | |
1086 | static inline bool skb_flow_dissect(const struct sk_buff *skb, | |
1087 | struct flow_dissector *flow_dissector, | |
cd79a238 | 1088 | void *target_container, unsigned int flags) |
e5276937 TH |
1089 | { |
1090 | return __skb_flow_dissect(skb, flow_dissector, target_container, | |
cd79a238 | 1091 | NULL, 0, 0, 0, flags); |
e5276937 TH |
1092 | } |
1093 | ||
1094 | static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb, | |
cd79a238 TH |
1095 | struct flow_keys *flow, |
1096 | unsigned int flags) | |
e5276937 TH |
1097 | { |
1098 | memset(flow, 0, sizeof(*flow)); | |
1099 | return __skb_flow_dissect(skb, &flow_keys_dissector, flow, | |
cd79a238 | 1100 | NULL, 0, 0, 0, flags); |
e5276937 TH |
1101 | } |
1102 | ||
1103 | static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow, | |
1104 | void *data, __be16 proto, | |
cd79a238 TH |
1105 | int nhoff, int hlen, |
1106 | unsigned int flags) | |
e5276937 TH |
1107 | { |
1108 | memset(flow, 0, sizeof(*flow)); | |
1109 | return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow, | |
cd79a238 | 1110 | data, proto, nhoff, hlen, flags); |
e5276937 TH |
1111 | } |
1112 | ||
3958afa1 | 1113 | static inline __u32 skb_get_hash(struct sk_buff *skb) |
bfb564e7 | 1114 | { |
a3b18ddb | 1115 | if (!skb->l4_hash && !skb->sw_hash) |
3958afa1 | 1116 | __skb_get_hash(skb); |
bfb564e7 | 1117 | |
61b905da | 1118 | return skb->hash; |
bfb564e7 KK |
1119 | } |
1120 | ||
20a17bf6 | 1121 | __u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6); |
f70ea018 | 1122 | |
20a17bf6 | 1123 | static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6) |
f70ea018 | 1124 | { |
c6cc1ca7 TH |
1125 | if (!skb->l4_hash && !skb->sw_hash) { |
1126 | struct flow_keys keys; | |
de4c1f8b | 1127 | __u32 hash = __get_hash_from_flowi6(fl6, &keys); |
c6cc1ca7 | 1128 | |
de4c1f8b | 1129 | __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys)); |
c6cc1ca7 | 1130 | } |
f70ea018 TH |
1131 | |
1132 | return skb->hash; | |
1133 | } | |
1134 | ||
20a17bf6 | 1135 | __u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl); |
f70ea018 | 1136 | |
20a17bf6 | 1137 | static inline __u32 skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4) |
f70ea018 | 1138 | { |
c6cc1ca7 TH |
1139 | if (!skb->l4_hash && !skb->sw_hash) { |
1140 | struct flow_keys keys; | |
de4c1f8b | 1141 | __u32 hash = __get_hash_from_flowi4(fl4, &keys); |
c6cc1ca7 | 1142 | |
de4c1f8b | 1143 | __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys)); |
c6cc1ca7 | 1144 | } |
f70ea018 TH |
1145 | |
1146 | return skb->hash; | |
1147 | } | |
1148 | ||
50fb7992 TH |
1149 | __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb); |
1150 | ||
57bdf7f4 TH |
1151 | static inline __u32 skb_get_hash_raw(const struct sk_buff *skb) |
1152 | { | |
61b905da | 1153 | return skb->hash; |
57bdf7f4 TH |
1154 | } |
1155 | ||
3df7a74e TH |
1156 | static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from) |
1157 | { | |
61b905da | 1158 | to->hash = from->hash; |
a3b18ddb | 1159 | to->sw_hash = from->sw_hash; |
61b905da | 1160 | to->l4_hash = from->l4_hash; |
3df7a74e TH |
1161 | }; |
1162 | ||
c29390c6 ED |
1163 | static inline void skb_sender_cpu_clear(struct sk_buff *skb) |
1164 | { | |
c29390c6 ED |
1165 | } |
1166 | ||
4305b541 ACM |
1167 | #ifdef NET_SKBUFF_DATA_USES_OFFSET |
1168 | static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) | |
1169 | { | |
1170 | return skb->head + skb->end; | |
1171 | } | |
ec47ea82 AD |
1172 | |
1173 | static inline unsigned int skb_end_offset(const struct sk_buff *skb) | |
1174 | { | |
1175 | return skb->end; | |
1176 | } | |
4305b541 ACM |
1177 | #else |
1178 | static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) | |
1179 | { | |
1180 | return skb->end; | |
1181 | } | |
ec47ea82 AD |
1182 | |
1183 | static inline unsigned int skb_end_offset(const struct sk_buff *skb) | |
1184 | { | |
1185 | return skb->end - skb->head; | |
1186 | } | |
4305b541 ACM |
1187 | #endif |
1188 | ||
1da177e4 | 1189 | /* Internal */ |
4305b541 | 1190 | #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB))) |
1da177e4 | 1191 | |
ac45f602 PO |
1192 | static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb) |
1193 | { | |
1194 | return &skb_shinfo(skb)->hwtstamps; | |
1195 | } | |
1196 | ||
1da177e4 LT |
1197 | /** |
1198 | * skb_queue_empty - check if a queue is empty | |
1199 | * @list: queue head | |
1200 | * | |
1201 | * Returns true if the queue is empty, false otherwise. | |
1202 | */ | |
1203 | static inline int skb_queue_empty(const struct sk_buff_head *list) | |
1204 | { | |
fd44b93c | 1205 | return list->next == (const struct sk_buff *) list; |
1da177e4 LT |
1206 | } |
1207 | ||
fc7ebb21 DM |
1208 | /** |
1209 | * skb_queue_is_last - check if skb is the last entry in the queue | |
1210 | * @list: queue head | |
1211 | * @skb: buffer | |
1212 | * | |
1213 | * Returns true if @skb is the last buffer on the list. | |
1214 | */ | |
1215 | static inline bool skb_queue_is_last(const struct sk_buff_head *list, | |
1216 | const struct sk_buff *skb) | |
1217 | { | |
fd44b93c | 1218 | return skb->next == (const struct sk_buff *) list; |
fc7ebb21 DM |
1219 | } |
1220 | ||
832d11c5 IJ |
1221 | /** |
1222 | * skb_queue_is_first - check if skb is the first entry in the queue | |
1223 | * @list: queue head | |
1224 | * @skb: buffer | |
1225 | * | |
1226 | * Returns true if @skb is the first buffer on the list. | |
1227 | */ | |
1228 | static inline bool skb_queue_is_first(const struct sk_buff_head *list, | |
1229 | const struct sk_buff *skb) | |
1230 | { | |
fd44b93c | 1231 | return skb->prev == (const struct sk_buff *) list; |
832d11c5 IJ |
1232 | } |
1233 | ||
249c8b42 DM |
1234 | /** |
1235 | * skb_queue_next - return the next packet in the queue | |
1236 | * @list: queue head | |
1237 | * @skb: current buffer | |
1238 | * | |
1239 | * Return the next packet in @list after @skb. It is only valid to | |
1240 | * call this if skb_queue_is_last() evaluates to false. | |
1241 | */ | |
1242 | static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list, | |
1243 | const struct sk_buff *skb) | |
1244 | { | |
1245 | /* This BUG_ON may seem severe, but if we just return then we | |
1246 | * are going to dereference garbage. | |
1247 | */ | |
1248 | BUG_ON(skb_queue_is_last(list, skb)); | |
1249 | return skb->next; | |
1250 | } | |
1251 | ||
832d11c5 IJ |
1252 | /** |
1253 | * skb_queue_prev - return the prev packet in the queue | |
1254 | * @list: queue head | |
1255 | * @skb: current buffer | |
1256 | * | |
1257 | * Return the prev packet in @list before @skb. It is only valid to | |
1258 | * call this if skb_queue_is_first() evaluates to false. | |
1259 | */ | |
1260 | static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list, | |
1261 | const struct sk_buff *skb) | |
1262 | { | |
1263 | /* This BUG_ON may seem severe, but if we just return then we | |
1264 | * are going to dereference garbage. | |
1265 | */ | |
1266 | BUG_ON(skb_queue_is_first(list, skb)); | |
1267 | return skb->prev; | |
1268 | } | |
1269 | ||
1da177e4 LT |
1270 | /** |
1271 | * skb_get - reference buffer | |
1272 | * @skb: buffer to reference | |
1273 | * | |
1274 | * Makes another reference to a socket buffer and returns a pointer | |
1275 | * to the buffer. | |
1276 | */ | |
1277 | static inline struct sk_buff *skb_get(struct sk_buff *skb) | |
1278 | { | |
1279 | atomic_inc(&skb->users); | |
1280 | return skb; | |
1281 | } | |
1282 | ||
1283 | /* | |
1284 | * If users == 1, we are the only owner and are can avoid redundant | |
1285 | * atomic change. | |
1286 | */ | |
1287 | ||
1da177e4 LT |
1288 | /** |
1289 | * skb_cloned - is the buffer a clone | |
1290 | * @skb: buffer to check | |
1291 | * | |
1292 | * Returns true if the buffer was generated with skb_clone() and is | |
1293 | * one of multiple shared copies of the buffer. Cloned buffers are | |
1294 | * shared data so must not be written to under normal circumstances. | |
1295 | */ | |
1296 | static inline int skb_cloned(const struct sk_buff *skb) | |
1297 | { | |
1298 | return skb->cloned && | |
1299 | (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1; | |
1300 | } | |
1301 | ||
14bbd6a5 PS |
1302 | static inline int skb_unclone(struct sk_buff *skb, gfp_t pri) |
1303 | { | |
d0164adc | 1304 | might_sleep_if(gfpflags_allow_blocking(pri)); |
14bbd6a5 PS |
1305 | |
1306 | if (skb_cloned(skb)) | |
1307 | return pskb_expand_head(skb, 0, 0, pri); | |
1308 | ||
1309 | return 0; | |
1310 | } | |
1311 | ||
1da177e4 LT |
1312 | /** |
1313 | * skb_header_cloned - is the header a clone | |
1314 | * @skb: buffer to check | |
1315 | * | |
1316 | * Returns true if modifying the header part of the buffer requires | |
1317 | * the data to be copied. | |
1318 | */ | |
1319 | static inline int skb_header_cloned(const struct sk_buff *skb) | |
1320 | { | |
1321 | int dataref; | |
1322 | ||
1323 | if (!skb->cloned) | |
1324 | return 0; | |
1325 | ||
1326 | dataref = atomic_read(&skb_shinfo(skb)->dataref); | |
1327 | dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT); | |
1328 | return dataref != 1; | |
1329 | } | |
1330 | ||
1331 | /** | |
1332 | * skb_header_release - release reference to header | |
1333 | * @skb: buffer to operate on | |
1334 | * | |
1335 | * Drop a reference to the header part of the buffer. This is done | |
1336 | * by acquiring a payload reference. You must not read from the header | |
1337 | * part of skb->data after this. | |
f4a775d1 | 1338 | * Note : Check if you can use __skb_header_release() instead. |
1da177e4 LT |
1339 | */ |
1340 | static inline void skb_header_release(struct sk_buff *skb) | |
1341 | { | |
1342 | BUG_ON(skb->nohdr); | |
1343 | skb->nohdr = 1; | |
1344 | atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref); | |
1345 | } | |
1346 | ||
f4a775d1 ED |
1347 | /** |
1348 | * __skb_header_release - release reference to header | |
1349 | * @skb: buffer to operate on | |
1350 | * | |
1351 | * Variant of skb_header_release() assuming skb is private to caller. | |
1352 | * We can avoid one atomic operation. | |
1353 | */ | |
1354 | static inline void __skb_header_release(struct sk_buff *skb) | |
1355 | { | |
1356 | skb->nohdr = 1; | |
1357 | atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT)); | |
1358 | } | |
1359 | ||
1360 | ||
1da177e4 LT |
1361 | /** |
1362 | * skb_shared - is the buffer shared | |
1363 | * @skb: buffer to check | |
1364 | * | |
1365 | * Returns true if more than one person has a reference to this | |
1366 | * buffer. | |
1367 | */ | |
1368 | static inline int skb_shared(const struct sk_buff *skb) | |
1369 | { | |
1370 | return atomic_read(&skb->users) != 1; | |
1371 | } | |
1372 | ||
1373 | /** | |
1374 | * skb_share_check - check if buffer is shared and if so clone it | |
1375 | * @skb: buffer to check | |
1376 | * @pri: priority for memory allocation | |
1377 | * | |
1378 | * If the buffer is shared the buffer is cloned and the old copy | |
1379 | * drops a reference. A new clone with a single reference is returned. | |
1380 | * If the buffer is not shared the original buffer is returned. When | |
1381 | * being called from interrupt status or with spinlocks held pri must | |
1382 | * be GFP_ATOMIC. | |
1383 | * | |
1384 | * NULL is returned on a memory allocation failure. | |
1385 | */ | |
47061bc4 | 1386 | static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri) |
1da177e4 | 1387 | { |
d0164adc | 1388 | might_sleep_if(gfpflags_allow_blocking(pri)); |
1da177e4 LT |
1389 | if (skb_shared(skb)) { |
1390 | struct sk_buff *nskb = skb_clone(skb, pri); | |
47061bc4 ED |
1391 | |
1392 | if (likely(nskb)) | |
1393 | consume_skb(skb); | |
1394 | else | |
1395 | kfree_skb(skb); | |
1da177e4 LT |
1396 | skb = nskb; |
1397 | } | |
1398 | return skb; | |
1399 | } | |
1400 | ||
1401 | /* | |
1402 | * Copy shared buffers into a new sk_buff. We effectively do COW on | |
1403 | * packets to handle cases where we have a local reader and forward | |
1404 | * and a couple of other messy ones. The normal one is tcpdumping | |
1405 | * a packet thats being forwarded. | |
1406 | */ | |
1407 | ||
1408 | /** | |
1409 | * skb_unshare - make a copy of a shared buffer | |
1410 | * @skb: buffer to check | |
1411 | * @pri: priority for memory allocation | |
1412 | * | |
1413 | * If the socket buffer is a clone then this function creates a new | |
1414 | * copy of the data, drops a reference count on the old copy and returns | |
1415 | * the new copy with the reference count at 1. If the buffer is not a clone | |
1416 | * the original buffer is returned. When called with a spinlock held or | |
1417 | * from interrupt state @pri must be %GFP_ATOMIC | |
1418 | * | |
1419 | * %NULL is returned on a memory allocation failure. | |
1420 | */ | |
e2bf521d | 1421 | static inline struct sk_buff *skb_unshare(struct sk_buff *skb, |
dd0fc66f | 1422 | gfp_t pri) |
1da177e4 | 1423 | { |
d0164adc | 1424 | might_sleep_if(gfpflags_allow_blocking(pri)); |
1da177e4 LT |
1425 | if (skb_cloned(skb)) { |
1426 | struct sk_buff *nskb = skb_copy(skb, pri); | |
31eff81e AA |
1427 | |
1428 | /* Free our shared copy */ | |
1429 | if (likely(nskb)) | |
1430 | consume_skb(skb); | |
1431 | else | |
1432 | kfree_skb(skb); | |
1da177e4 LT |
1433 | skb = nskb; |
1434 | } | |
1435 | return skb; | |
1436 | } | |
1437 | ||
1438 | /** | |
1a5778aa | 1439 | * skb_peek - peek at the head of an &sk_buff_head |
1da177e4 LT |
1440 | * @list_: list to peek at |
1441 | * | |
1442 | * Peek an &sk_buff. Unlike most other operations you _MUST_ | |
1443 | * be careful with this one. A peek leaves the buffer on the | |
1444 | * list and someone else may run off with it. You must hold | |
1445 | * the appropriate locks or have a private queue to do this. | |
1446 | * | |
1447 | * Returns %NULL for an empty list or a pointer to the head element. | |
1448 | * The reference count is not incremented and the reference is therefore | |
1449 | * volatile. Use with caution. | |
1450 | */ | |
05bdd2f1 | 1451 | static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_) |
1da177e4 | 1452 | { |
18d07000 ED |
1453 | struct sk_buff *skb = list_->next; |
1454 | ||
1455 | if (skb == (struct sk_buff *)list_) | |
1456 | skb = NULL; | |
1457 | return skb; | |
1da177e4 LT |
1458 | } |
1459 | ||
da5ef6e5 PE |
1460 | /** |
1461 | * skb_peek_next - peek skb following the given one from a queue | |
1462 | * @skb: skb to start from | |
1463 | * @list_: list to peek at | |
1464 | * | |
1465 | * Returns %NULL when the end of the list is met or a pointer to the | |
1466 | * next element. The reference count is not incremented and the | |
1467 | * reference is therefore volatile. Use with caution. | |
1468 | */ | |
1469 | static inline struct sk_buff *skb_peek_next(struct sk_buff *skb, | |
1470 | const struct sk_buff_head *list_) | |
1471 | { | |
1472 | struct sk_buff *next = skb->next; | |
18d07000 | 1473 | |
da5ef6e5 PE |
1474 | if (next == (struct sk_buff *)list_) |
1475 | next = NULL; | |
1476 | return next; | |
1477 | } | |
1478 | ||
1da177e4 | 1479 | /** |
1a5778aa | 1480 | * skb_peek_tail - peek at the tail of an &sk_buff_head |
1da177e4 LT |
1481 | * @list_: list to peek at |
1482 | * | |
1483 | * Peek an &sk_buff. Unlike most other operations you _MUST_ | |
1484 | * be careful with this one. A peek leaves the buffer on the | |
1485 | * list and someone else may run off with it. You must hold | |
1486 | * the appropriate locks or have a private queue to do this. | |
1487 | * | |
1488 | * Returns %NULL for an empty list or a pointer to the tail element. | |
1489 | * The reference count is not incremented and the reference is therefore | |
1490 | * volatile. Use with caution. | |
1491 | */ | |
05bdd2f1 | 1492 | static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_) |
1da177e4 | 1493 | { |
18d07000 ED |
1494 | struct sk_buff *skb = list_->prev; |
1495 | ||
1496 | if (skb == (struct sk_buff *)list_) | |
1497 | skb = NULL; | |
1498 | return skb; | |
1499 | ||
1da177e4 LT |
1500 | } |
1501 | ||
1502 | /** | |
1503 | * skb_queue_len - get queue length | |
1504 | * @list_: list to measure | |
1505 | * | |
1506 | * Return the length of an &sk_buff queue. | |
1507 | */ | |
1508 | static inline __u32 skb_queue_len(const struct sk_buff_head *list_) | |
1509 | { | |
1510 | return list_->qlen; | |
1511 | } | |
1512 | ||
67fed459 DM |
1513 | /** |
1514 | * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head | |
1515 | * @list: queue to initialize | |
1516 | * | |
1517 | * This initializes only the list and queue length aspects of | |
1518 | * an sk_buff_head object. This allows to initialize the list | |
1519 | * aspects of an sk_buff_head without reinitializing things like | |
1520 | * the spinlock. It can also be used for on-stack sk_buff_head | |
1521 | * objects where the spinlock is known to not be used. | |
1522 | */ | |
1523 | static inline void __skb_queue_head_init(struct sk_buff_head *list) | |
1524 | { | |
1525 | list->prev = list->next = (struct sk_buff *)list; | |
1526 | list->qlen = 0; | |
1527 | } | |
1528 | ||
76f10ad0 AV |
1529 | /* |
1530 | * This function creates a split out lock class for each invocation; | |
1531 | * this is needed for now since a whole lot of users of the skb-queue | |
1532 | * infrastructure in drivers have different locking usage (in hardirq) | |
1533 | * than the networking core (in softirq only). In the long run either the | |
1534 | * network layer or drivers should need annotation to consolidate the | |
1535 | * main types of usage into 3 classes. | |
1536 | */ | |
1da177e4 LT |
1537 | static inline void skb_queue_head_init(struct sk_buff_head *list) |
1538 | { | |
1539 | spin_lock_init(&list->lock); | |
67fed459 | 1540 | __skb_queue_head_init(list); |
1da177e4 LT |
1541 | } |
1542 | ||
c2ecba71 PE |
1543 | static inline void skb_queue_head_init_class(struct sk_buff_head *list, |
1544 | struct lock_class_key *class) | |
1545 | { | |
1546 | skb_queue_head_init(list); | |
1547 | lockdep_set_class(&list->lock, class); | |
1548 | } | |
1549 | ||
1da177e4 | 1550 | /* |
bf299275 | 1551 | * Insert an sk_buff on a list. |
1da177e4 LT |
1552 | * |
1553 | * The "__skb_xxxx()" functions are the non-atomic ones that | |
1554 | * can only be called with interrupts disabled. | |
1555 | */ | |
7965bd4d JP |
1556 | void skb_insert(struct sk_buff *old, struct sk_buff *newsk, |
1557 | struct sk_buff_head *list); | |
bf299275 GR |
1558 | static inline void __skb_insert(struct sk_buff *newsk, |
1559 | struct sk_buff *prev, struct sk_buff *next, | |
1560 | struct sk_buff_head *list) | |
1561 | { | |
1562 | newsk->next = next; | |
1563 | newsk->prev = prev; | |
1564 | next->prev = prev->next = newsk; | |
1565 | list->qlen++; | |
1566 | } | |
1da177e4 | 1567 | |
67fed459 DM |
1568 | static inline void __skb_queue_splice(const struct sk_buff_head *list, |
1569 | struct sk_buff *prev, | |
1570 | struct sk_buff *next) | |
1571 | { | |
1572 | struct sk_buff *first = list->next; | |
1573 | struct sk_buff *last = list->prev; | |
1574 | ||
1575 | first->prev = prev; | |
1576 | prev->next = first; | |
1577 | ||
1578 | last->next = next; | |
1579 | next->prev = last; | |
1580 | } | |
1581 | ||
1582 | /** | |
1583 | * skb_queue_splice - join two skb lists, this is designed for stacks | |
1584 | * @list: the new list to add | |
1585 | * @head: the place to add it in the first list | |
1586 | */ | |
1587 | static inline void skb_queue_splice(const struct sk_buff_head *list, | |
1588 | struct sk_buff_head *head) | |
1589 | { | |
1590 | if (!skb_queue_empty(list)) { | |
1591 | __skb_queue_splice(list, (struct sk_buff *) head, head->next); | |
1d4a31dd | 1592 | head->qlen += list->qlen; |
67fed459 DM |
1593 | } |
1594 | } | |
1595 | ||
1596 | /** | |
d9619496 | 1597 | * skb_queue_splice_init - join two skb lists and reinitialise the emptied list |
67fed459 DM |
1598 | * @list: the new list to add |
1599 | * @head: the place to add it in the first list | |
1600 | * | |
1601 | * The list at @list is reinitialised | |
1602 | */ | |
1603 | static inline void skb_queue_splice_init(struct sk_buff_head *list, | |
1604 | struct sk_buff_head *head) | |
1605 | { | |
1606 | if (!skb_queue_empty(list)) { | |
1607 | __skb_queue_splice(list, (struct sk_buff *) head, head->next); | |
1d4a31dd | 1608 | head->qlen += list->qlen; |
67fed459 DM |
1609 | __skb_queue_head_init(list); |
1610 | } | |
1611 | } | |
1612 | ||
1613 | /** | |
1614 | * skb_queue_splice_tail - join two skb lists, each list being a queue | |
1615 | * @list: the new list to add | |
1616 | * @head: the place to add it in the first list | |
1617 | */ | |
1618 | static inline void skb_queue_splice_tail(const struct sk_buff_head *list, | |
1619 | struct sk_buff_head *head) | |
1620 | { | |
1621 | if (!skb_queue_empty(list)) { | |
1622 | __skb_queue_splice(list, head->prev, (struct sk_buff *) head); | |
1d4a31dd | 1623 | head->qlen += list->qlen; |
67fed459 DM |
1624 | } |
1625 | } | |
1626 | ||
1627 | /** | |
d9619496 | 1628 | * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list |
67fed459 DM |
1629 | * @list: the new list to add |
1630 | * @head: the place to add it in the first list | |
1631 | * | |
1632 | * Each of the lists is a queue. | |
1633 | * The list at @list is reinitialised | |
1634 | */ | |
1635 | static inline void skb_queue_splice_tail_init(struct sk_buff_head *list, | |
1636 | struct sk_buff_head *head) | |
1637 | { | |
1638 | if (!skb_queue_empty(list)) { | |
1639 | __skb_queue_splice(list, head->prev, (struct sk_buff *) head); | |
1d4a31dd | 1640 | head->qlen += list->qlen; |
67fed459 DM |
1641 | __skb_queue_head_init(list); |
1642 | } | |
1643 | } | |
1644 | ||
1da177e4 | 1645 | /** |
300ce174 | 1646 | * __skb_queue_after - queue a buffer at the list head |
1da177e4 | 1647 | * @list: list to use |
300ce174 | 1648 | * @prev: place after this buffer |
1da177e4 LT |
1649 | * @newsk: buffer to queue |
1650 | * | |
300ce174 | 1651 | * Queue a buffer int the middle of a list. This function takes no locks |
1da177e4 LT |
1652 | * and you must therefore hold required locks before calling it. |
1653 | * | |
1654 | * A buffer cannot be placed on two lists at the same time. | |
1655 | */ | |
300ce174 SH |
1656 | static inline void __skb_queue_after(struct sk_buff_head *list, |
1657 | struct sk_buff *prev, | |
1658 | struct sk_buff *newsk) | |
1da177e4 | 1659 | { |
bf299275 | 1660 | __skb_insert(newsk, prev, prev->next, list); |
1da177e4 LT |
1661 | } |
1662 | ||
7965bd4d JP |
1663 | void skb_append(struct sk_buff *old, struct sk_buff *newsk, |
1664 | struct sk_buff_head *list); | |
7de6c033 | 1665 | |
f5572855 GR |
1666 | static inline void __skb_queue_before(struct sk_buff_head *list, |
1667 | struct sk_buff *next, | |
1668 | struct sk_buff *newsk) | |
1669 | { | |
1670 | __skb_insert(newsk, next->prev, next, list); | |
1671 | } | |
1672 | ||
300ce174 SH |
1673 | /** |
1674 | * __skb_queue_head - queue a buffer at the list head | |
1675 | * @list: list to use | |
1676 | * @newsk: buffer to queue | |
1677 | * | |
1678 | * Queue a buffer at the start of a list. This function takes no locks | |
1679 | * and you must therefore hold required locks before calling it. | |
1680 | * | |
1681 | * A buffer cannot be placed on two lists at the same time. | |
1682 | */ | |
7965bd4d | 1683 | void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk); |
300ce174 SH |
1684 | static inline void __skb_queue_head(struct sk_buff_head *list, |
1685 | struct sk_buff *newsk) | |
1686 | { | |
1687 | __skb_queue_after(list, (struct sk_buff *)list, newsk); | |
1688 | } | |
1689 | ||
1da177e4 LT |
1690 | /** |
1691 | * __skb_queue_tail - queue a buffer at the list tail | |
1692 | * @list: list to use | |
1693 | * @newsk: buffer to queue | |
1694 | * | |
1695 | * Queue a buffer at the end of a list. This function takes no locks | |
1696 | * and you must therefore hold required locks before calling it. | |
1697 | * | |
1698 | * A buffer cannot be placed on two lists at the same time. | |
1699 | */ | |
7965bd4d | 1700 | void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk); |
1da177e4 LT |
1701 | static inline void __skb_queue_tail(struct sk_buff_head *list, |
1702 | struct sk_buff *newsk) | |
1703 | { | |
f5572855 | 1704 | __skb_queue_before(list, (struct sk_buff *)list, newsk); |
1da177e4 LT |
1705 | } |
1706 | ||
1da177e4 LT |
1707 | /* |
1708 | * remove sk_buff from list. _Must_ be called atomically, and with | |
1709 | * the list known.. | |
1710 | */ | |
7965bd4d | 1711 | void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list); |
1da177e4 LT |
1712 | static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list) |
1713 | { | |
1714 | struct sk_buff *next, *prev; | |
1715 | ||
1716 | list->qlen--; | |
1717 | next = skb->next; | |
1718 | prev = skb->prev; | |
1719 | skb->next = skb->prev = NULL; | |
1da177e4 LT |
1720 | next->prev = prev; |
1721 | prev->next = next; | |
1722 | } | |
1723 | ||
f525c06d GR |
1724 | /** |
1725 | * __skb_dequeue - remove from the head of the queue | |
1726 | * @list: list to dequeue from | |
1727 | * | |
1728 | * Remove the head of the list. This function does not take any locks | |
1729 | * so must be used with appropriate locks held only. The head item is | |
1730 | * returned or %NULL if the list is empty. | |
1731 | */ | |
7965bd4d | 1732 | struct sk_buff *skb_dequeue(struct sk_buff_head *list); |
f525c06d GR |
1733 | static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list) |
1734 | { | |
1735 | struct sk_buff *skb = skb_peek(list); | |
1736 | if (skb) | |
1737 | __skb_unlink(skb, list); | |
1738 | return skb; | |
1739 | } | |
1da177e4 LT |
1740 | |
1741 | /** | |
1742 | * __skb_dequeue_tail - remove from the tail of the queue | |
1743 | * @list: list to dequeue from | |
1744 | * | |
1745 | * Remove the tail of the list. This function does not take any locks | |
1746 | * so must be used with appropriate locks held only. The tail item is | |
1747 | * returned or %NULL if the list is empty. | |
1748 | */ | |
7965bd4d | 1749 | struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list); |
1da177e4 LT |
1750 | static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list) |
1751 | { | |
1752 | struct sk_buff *skb = skb_peek_tail(list); | |
1753 | if (skb) | |
1754 | __skb_unlink(skb, list); | |
1755 | return skb; | |
1756 | } | |
1757 | ||
1758 | ||
bdcc0924 | 1759 | static inline bool skb_is_nonlinear(const struct sk_buff *skb) |
1da177e4 LT |
1760 | { |
1761 | return skb->data_len; | |
1762 | } | |
1763 | ||
1764 | static inline unsigned int skb_headlen(const struct sk_buff *skb) | |
1765 | { | |
1766 | return skb->len - skb->data_len; | |
1767 | } | |
1768 | ||
1769 | static inline int skb_pagelen(const struct sk_buff *skb) | |
1770 | { | |
1771 | int i, len = 0; | |
1772 | ||
1773 | for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--) | |
9e903e08 | 1774 | len += skb_frag_size(&skb_shinfo(skb)->frags[i]); |
1da177e4 LT |
1775 | return len + skb_headlen(skb); |
1776 | } | |
1777 | ||
131ea667 IC |
1778 | /** |
1779 | * __skb_fill_page_desc - initialise a paged fragment in an skb | |
1780 | * @skb: buffer containing fragment to be initialised | |
1781 | * @i: paged fragment index to initialise | |
1782 | * @page: the page to use for this fragment | |
1783 | * @off: the offset to the data with @page | |
1784 | * @size: the length of the data | |
1785 | * | |
1786 | * Initialises the @i'th fragment of @skb to point to &size bytes at | |
1787 | * offset @off within @page. | |
1788 | * | |
1789 | * Does not take any additional reference on the fragment. | |
1790 | */ | |
1791 | static inline void __skb_fill_page_desc(struct sk_buff *skb, int i, | |
1792 | struct page *page, int off, int size) | |
1da177e4 LT |
1793 | { |
1794 | skb_frag_t *frag = &skb_shinfo(skb)->frags[i]; | |
1795 | ||
c48a11c7 | 1796 | /* |
2f064f34 MH |
1797 | * Propagate page pfmemalloc to the skb if we can. The problem is |
1798 | * that not all callers have unique ownership of the page but rely | |
1799 | * on page_is_pfmemalloc doing the right thing(tm). | |
c48a11c7 | 1800 | */ |
a8605c60 | 1801 | frag->page.p = page; |
1da177e4 | 1802 | frag->page_offset = off; |
9e903e08 | 1803 | skb_frag_size_set(frag, size); |
cca7af38 PE |
1804 | |
1805 | page = compound_head(page); | |
2f064f34 | 1806 | if (page_is_pfmemalloc(page)) |
cca7af38 | 1807 | skb->pfmemalloc = true; |
131ea667 IC |
1808 | } |
1809 | ||
1810 | /** | |
1811 | * skb_fill_page_desc - initialise a paged fragment in an skb | |
1812 | * @skb: buffer containing fragment to be initialised | |
1813 | * @i: paged fragment index to initialise | |
1814 | * @page: the page to use for this fragment | |
1815 | * @off: the offset to the data with @page | |
1816 | * @size: the length of the data | |
1817 | * | |
1818 | * As per __skb_fill_page_desc() -- initialises the @i'th fragment of | |
bc32383c | 1819 | * @skb to point to @size bytes at offset @off within @page. In |
131ea667 IC |
1820 | * addition updates @skb such that @i is the last fragment. |
1821 | * | |
1822 | * Does not take any additional reference on the fragment. | |
1823 | */ | |
1824 | static inline void skb_fill_page_desc(struct sk_buff *skb, int i, | |
1825 | struct page *page, int off, int size) | |
1826 | { | |
1827 | __skb_fill_page_desc(skb, i, page, off, size); | |
1da177e4 LT |
1828 | skb_shinfo(skb)->nr_frags = i + 1; |
1829 | } | |
1830 | ||
7965bd4d JP |
1831 | void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off, |
1832 | int size, unsigned int truesize); | |
654bed16 | 1833 | |
f8e617e1 JW |
1834 | void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size, |
1835 | unsigned int truesize); | |
1836 | ||
1da177e4 | 1837 | #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags) |
21dc3301 | 1838 | #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb)) |
1da177e4 LT |
1839 | #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb)) |
1840 | ||
27a884dc ACM |
1841 | #ifdef NET_SKBUFF_DATA_USES_OFFSET |
1842 | static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) | |
1843 | { | |
1844 | return skb->head + skb->tail; | |
1845 | } | |
1846 | ||
1847 | static inline void skb_reset_tail_pointer(struct sk_buff *skb) | |
1848 | { | |
1849 | skb->tail = skb->data - skb->head; | |
1850 | } | |
1851 | ||
1852 | static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) | |
1853 | { | |
1854 | skb_reset_tail_pointer(skb); | |
1855 | skb->tail += offset; | |
1856 | } | |
7cc46190 | 1857 | |
27a884dc ACM |
1858 | #else /* NET_SKBUFF_DATA_USES_OFFSET */ |
1859 | static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) | |
1860 | { | |
1861 | return skb->tail; | |
1862 | } | |
1863 | ||
1864 | static inline void skb_reset_tail_pointer(struct sk_buff *skb) | |
1865 | { | |
1866 | skb->tail = skb->data; | |
1867 | } | |
1868 | ||
1869 | static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) | |
1870 | { | |
1871 | skb->tail = skb->data + offset; | |
1872 | } | |
4305b541 | 1873 | |
27a884dc ACM |
1874 | #endif /* NET_SKBUFF_DATA_USES_OFFSET */ |
1875 | ||
1da177e4 LT |
1876 | /* |
1877 | * Add data to an sk_buff | |
1878 | */ | |
0c7ddf36 | 1879 | unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len); |
7965bd4d | 1880 | unsigned char *skb_put(struct sk_buff *skb, unsigned int len); |
1da177e4 LT |
1881 | static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len) |
1882 | { | |
27a884dc | 1883 | unsigned char *tmp = skb_tail_pointer(skb); |
1da177e4 LT |
1884 | SKB_LINEAR_ASSERT(skb); |
1885 | skb->tail += len; | |
1886 | skb->len += len; | |
1887 | return tmp; | |
1888 | } | |
1889 | ||
7965bd4d | 1890 | unsigned char *skb_push(struct sk_buff *skb, unsigned int len); |
1da177e4 LT |
1891 | static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len) |
1892 | { | |
1893 | skb->data -= len; | |
1894 | skb->len += len; | |
1895 | return skb->data; | |
1896 | } | |
1897 | ||
7965bd4d | 1898 | unsigned char *skb_pull(struct sk_buff *skb, unsigned int len); |
1da177e4 LT |
1899 | static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len) |
1900 | { | |
1901 | skb->len -= len; | |
1902 | BUG_ON(skb->len < skb->data_len); | |
1903 | return skb->data += len; | |
1904 | } | |
1905 | ||
47d29646 DM |
1906 | static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len) |
1907 | { | |
1908 | return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len); | |
1909 | } | |
1910 | ||
7965bd4d | 1911 | unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta); |
1da177e4 LT |
1912 | |
1913 | static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len) | |
1914 | { | |
1915 | if (len > skb_headlen(skb) && | |
987c402a | 1916 | !__pskb_pull_tail(skb, len - skb_headlen(skb))) |
1da177e4 LT |
1917 | return NULL; |
1918 | skb->len -= len; | |
1919 | return skb->data += len; | |
1920 | } | |
1921 | ||
1922 | static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len) | |
1923 | { | |
1924 | return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len); | |
1925 | } | |
1926 | ||
1927 | static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len) | |
1928 | { | |
1929 | if (likely(len <= skb_headlen(skb))) | |
1930 | return 1; | |
1931 | if (unlikely(len > skb->len)) | |
1932 | return 0; | |
987c402a | 1933 | return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL; |
1da177e4 LT |
1934 | } |
1935 | ||
1936 | /** | |
1937 | * skb_headroom - bytes at buffer head | |
1938 | * @skb: buffer to check | |
1939 | * | |
1940 | * Return the number of bytes of free space at the head of an &sk_buff. | |
1941 | */ | |
c2636b4d | 1942 | static inline unsigned int skb_headroom(const struct sk_buff *skb) |
1da177e4 LT |
1943 | { |
1944 | return skb->data - skb->head; | |
1945 | } | |
1946 | ||
1947 | /** | |
1948 | * skb_tailroom - bytes at buffer end | |
1949 | * @skb: buffer to check | |
1950 | * | |
1951 | * Return the number of bytes of free space at the tail of an sk_buff | |
1952 | */ | |
1953 | static inline int skb_tailroom(const struct sk_buff *skb) | |
1954 | { | |
4305b541 | 1955 | return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail; |
1da177e4 LT |
1956 | } |
1957 | ||
a21d4572 ED |
1958 | /** |
1959 | * skb_availroom - bytes at buffer end | |
1960 | * @skb: buffer to check | |
1961 | * | |
1962 | * Return the number of bytes of free space at the tail of an sk_buff | |
1963 | * allocated by sk_stream_alloc() | |
1964 | */ | |
1965 | static inline int skb_availroom(const struct sk_buff *skb) | |
1966 | { | |
16fad69c ED |
1967 | if (skb_is_nonlinear(skb)) |
1968 | return 0; | |
1969 | ||
1970 | return skb->end - skb->tail - skb->reserved_tailroom; | |
a21d4572 ED |
1971 | } |
1972 | ||
1da177e4 LT |
1973 | /** |
1974 | * skb_reserve - adjust headroom | |
1975 | * @skb: buffer to alter | |
1976 | * @len: bytes to move | |
1977 | * | |
1978 | * Increase the headroom of an empty &sk_buff by reducing the tail | |
1979 | * room. This is only allowed for an empty buffer. | |
1980 | */ | |
8243126c | 1981 | static inline void skb_reserve(struct sk_buff *skb, int len) |
1da177e4 LT |
1982 | { |
1983 | skb->data += len; | |
1984 | skb->tail += len; | |
1985 | } | |
1986 | ||
8bce6d7d TH |
1987 | #define ENCAP_TYPE_ETHER 0 |
1988 | #define ENCAP_TYPE_IPPROTO 1 | |
1989 | ||
1990 | static inline void skb_set_inner_protocol(struct sk_buff *skb, | |
1991 | __be16 protocol) | |
1992 | { | |
1993 | skb->inner_protocol = protocol; | |
1994 | skb->inner_protocol_type = ENCAP_TYPE_ETHER; | |
1995 | } | |
1996 | ||
1997 | static inline void skb_set_inner_ipproto(struct sk_buff *skb, | |
1998 | __u8 ipproto) | |
1999 | { | |
2000 | skb->inner_ipproto = ipproto; | |
2001 | skb->inner_protocol_type = ENCAP_TYPE_IPPROTO; | |
2002 | } | |
2003 | ||
6a674e9c JG |
2004 | static inline void skb_reset_inner_headers(struct sk_buff *skb) |
2005 | { | |
aefbd2b3 | 2006 | skb->inner_mac_header = skb->mac_header; |
6a674e9c JG |
2007 | skb->inner_network_header = skb->network_header; |
2008 | skb->inner_transport_header = skb->transport_header; | |
2009 | } | |
2010 | ||
0b5c9db1 JP |
2011 | static inline void skb_reset_mac_len(struct sk_buff *skb) |
2012 | { | |
2013 | skb->mac_len = skb->network_header - skb->mac_header; | |
2014 | } | |
2015 | ||
6a674e9c JG |
2016 | static inline unsigned char *skb_inner_transport_header(const struct sk_buff |
2017 | *skb) | |
2018 | { | |
2019 | return skb->head + skb->inner_transport_header; | |
2020 | } | |
2021 | ||
55dc5a9f TH |
2022 | static inline int skb_inner_transport_offset(const struct sk_buff *skb) |
2023 | { | |
2024 | return skb_inner_transport_header(skb) - skb->data; | |
2025 | } | |
2026 | ||
6a674e9c JG |
2027 | static inline void skb_reset_inner_transport_header(struct sk_buff *skb) |
2028 | { | |
2029 | skb->inner_transport_header = skb->data - skb->head; | |
2030 | } | |
2031 | ||
2032 | static inline void skb_set_inner_transport_header(struct sk_buff *skb, | |
2033 | const int offset) | |
2034 | { | |
2035 | skb_reset_inner_transport_header(skb); | |
2036 | skb->inner_transport_header += offset; | |
2037 | } | |
2038 | ||
2039 | static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb) | |
2040 | { | |
2041 | return skb->head + skb->inner_network_header; | |
2042 | } | |
2043 | ||
2044 | static inline void skb_reset_inner_network_header(struct sk_buff *skb) | |
2045 | { | |
2046 | skb->inner_network_header = skb->data - skb->head; | |
2047 | } | |
2048 | ||
2049 | static inline void skb_set_inner_network_header(struct sk_buff *skb, | |
2050 | const int offset) | |
2051 | { | |
2052 | skb_reset_inner_network_header(skb); | |
2053 | skb->inner_network_header += offset; | |
2054 | } | |
2055 | ||
aefbd2b3 PS |
2056 | static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb) |
2057 | { | |
2058 | return skb->head + skb->inner_mac_header; | |
2059 | } | |
2060 | ||
2061 | static inline void skb_reset_inner_mac_header(struct sk_buff *skb) | |
2062 | { | |
2063 | skb->inner_mac_header = skb->data - skb->head; | |
2064 | } | |
2065 | ||
2066 | static inline void skb_set_inner_mac_header(struct sk_buff *skb, | |
2067 | const int offset) | |
2068 | { | |
2069 | skb_reset_inner_mac_header(skb); | |
2070 | skb->inner_mac_header += offset; | |
2071 | } | |
fda55eca ED |
2072 | static inline bool skb_transport_header_was_set(const struct sk_buff *skb) |
2073 | { | |
35d04610 | 2074 | return skb->transport_header != (typeof(skb->transport_header))~0U; |
fda55eca ED |
2075 | } |
2076 | ||
9c70220b ACM |
2077 | static inline unsigned char *skb_transport_header(const struct sk_buff *skb) |
2078 | { | |
2e07fa9c | 2079 | return skb->head + skb->transport_header; |
9c70220b ACM |
2080 | } |
2081 | ||
badff6d0 ACM |
2082 | static inline void skb_reset_transport_header(struct sk_buff *skb) |
2083 | { | |
2e07fa9c | 2084 | skb->transport_header = skb->data - skb->head; |
badff6d0 ACM |
2085 | } |
2086 | ||
967b05f6 ACM |
2087 | static inline void skb_set_transport_header(struct sk_buff *skb, |
2088 | const int offset) | |
2089 | { | |
2e07fa9c ACM |
2090 | skb_reset_transport_header(skb); |
2091 | skb->transport_header += offset; | |
ea2ae17d ACM |
2092 | } |
2093 | ||
d56f90a7 ACM |
2094 | static inline unsigned char *skb_network_header(const struct sk_buff *skb) |
2095 | { | |
2e07fa9c | 2096 | return skb->head + skb->network_header; |
d56f90a7 ACM |
2097 | } |
2098 | ||
c1d2bbe1 ACM |
2099 | static inline void skb_reset_network_header(struct sk_buff *skb) |
2100 | { | |
2e07fa9c | 2101 | skb->network_header = skb->data - skb->head; |
c1d2bbe1 ACM |
2102 | } |
2103 | ||
c14d2450 ACM |
2104 | static inline void skb_set_network_header(struct sk_buff *skb, const int offset) |
2105 | { | |
2e07fa9c ACM |
2106 | skb_reset_network_header(skb); |
2107 | skb->network_header += offset; | |
c14d2450 ACM |
2108 | } |
2109 | ||
2e07fa9c | 2110 | static inline unsigned char *skb_mac_header(const struct sk_buff *skb) |
bbe735e4 | 2111 | { |
2e07fa9c | 2112 | return skb->head + skb->mac_header; |
bbe735e4 ACM |
2113 | } |
2114 | ||
2e07fa9c | 2115 | static inline int skb_mac_header_was_set(const struct sk_buff *skb) |
cfe1fc77 | 2116 | { |
35d04610 | 2117 | return skb->mac_header != (typeof(skb->mac_header))~0U; |
2e07fa9c ACM |
2118 | } |
2119 | ||
2120 | static inline void skb_reset_mac_header(struct sk_buff *skb) | |
2121 | { | |
2122 | skb->mac_header = skb->data - skb->head; | |
2123 | } | |
2124 | ||
2125 | static inline void skb_set_mac_header(struct sk_buff *skb, const int offset) | |
2126 | { | |
2127 | skb_reset_mac_header(skb); | |
2128 | skb->mac_header += offset; | |
2129 | } | |
2130 | ||
0e3da5bb TT |
2131 | static inline void skb_pop_mac_header(struct sk_buff *skb) |
2132 | { | |
2133 | skb->mac_header = skb->network_header; | |
2134 | } | |
2135 | ||
fbbdb8f0 YX |
2136 | static inline void skb_probe_transport_header(struct sk_buff *skb, |
2137 | const int offset_hint) | |
2138 | { | |
2139 | struct flow_keys keys; | |
2140 | ||
2141 | if (skb_transport_header_was_set(skb)) | |
2142 | return; | |
cd79a238 | 2143 | else if (skb_flow_dissect_flow_keys(skb, &keys, 0)) |
42aecaa9 | 2144 | skb_set_transport_header(skb, keys.control.thoff); |
fbbdb8f0 YX |
2145 | else |
2146 | skb_set_transport_header(skb, offset_hint); | |
2147 | } | |
2148 | ||
03606895 ED |
2149 | static inline void skb_mac_header_rebuild(struct sk_buff *skb) |
2150 | { | |
2151 | if (skb_mac_header_was_set(skb)) { | |
2152 | const unsigned char *old_mac = skb_mac_header(skb); | |
2153 | ||
2154 | skb_set_mac_header(skb, -skb->mac_len); | |
2155 | memmove(skb_mac_header(skb), old_mac, skb->mac_len); | |
2156 | } | |
2157 | } | |
2158 | ||
04fb451e MM |
2159 | static inline int skb_checksum_start_offset(const struct sk_buff *skb) |
2160 | { | |
2161 | return skb->csum_start - skb_headroom(skb); | |
2162 | } | |
2163 | ||
2e07fa9c ACM |
2164 | static inline int skb_transport_offset(const struct sk_buff *skb) |
2165 | { | |
2166 | return skb_transport_header(skb) - skb->data; | |
2167 | } | |
2168 | ||
2169 | static inline u32 skb_network_header_len(const struct sk_buff *skb) | |
2170 | { | |
2171 | return skb->transport_header - skb->network_header; | |
2172 | } | |
2173 | ||
6a674e9c JG |
2174 | static inline u32 skb_inner_network_header_len(const struct sk_buff *skb) |
2175 | { | |
2176 | return skb->inner_transport_header - skb->inner_network_header; | |
2177 | } | |
2178 | ||
2e07fa9c ACM |
2179 | static inline int skb_network_offset(const struct sk_buff *skb) |
2180 | { | |
2181 | return skb_network_header(skb) - skb->data; | |
2182 | } | |
48d49d0c | 2183 | |
6a674e9c JG |
2184 | static inline int skb_inner_network_offset(const struct sk_buff *skb) |
2185 | { | |
2186 | return skb_inner_network_header(skb) - skb->data; | |
2187 | } | |
2188 | ||
f9599ce1 CG |
2189 | static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len) |
2190 | { | |
2191 | return pskb_may_pull(skb, skb_network_offset(skb) + len); | |
2192 | } | |
2193 | ||
1da177e4 LT |
2194 | /* |
2195 | * CPUs often take a performance hit when accessing unaligned memory | |
2196 | * locations. The actual performance hit varies, it can be small if the | |
2197 | * hardware handles it or large if we have to take an exception and fix it | |
2198 | * in software. | |
2199 | * | |
2200 | * Since an ethernet header is 14 bytes network drivers often end up with | |
2201 | * the IP header at an unaligned offset. The IP header can be aligned by | |
2202 | * shifting the start of the packet by 2 bytes. Drivers should do this | |
2203 | * with: | |
2204 | * | |
8660c124 | 2205 | * skb_reserve(skb, NET_IP_ALIGN); |
1da177e4 LT |
2206 | * |
2207 | * The downside to this alignment of the IP header is that the DMA is now | |
2208 | * unaligned. On some architectures the cost of an unaligned DMA is high | |
2209 | * and this cost outweighs the gains made by aligning the IP header. | |
8660c124 | 2210 | * |
1da177e4 LT |
2211 | * Since this trade off varies between architectures, we allow NET_IP_ALIGN |
2212 | * to be overridden. | |
2213 | */ | |
2214 | #ifndef NET_IP_ALIGN | |
2215 | #define NET_IP_ALIGN 2 | |
2216 | #endif | |
2217 | ||
025be81e AB |
2218 | /* |
2219 | * The networking layer reserves some headroom in skb data (via | |
2220 | * dev_alloc_skb). This is used to avoid having to reallocate skb data when | |
2221 | * the header has to grow. In the default case, if the header has to grow | |
d6301d3d | 2222 | * 32 bytes or less we avoid the reallocation. |
025be81e AB |
2223 | * |
2224 | * Unfortunately this headroom changes the DMA alignment of the resulting | |
2225 | * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive | |
2226 | * on some architectures. An architecture can override this value, | |
2227 | * perhaps setting it to a cacheline in size (since that will maintain | |
2228 | * cacheline alignment of the DMA). It must be a power of 2. | |
2229 | * | |
d6301d3d | 2230 | * Various parts of the networking layer expect at least 32 bytes of |
025be81e | 2231 | * headroom, you should not reduce this. |
5933dd2f ED |
2232 | * |
2233 | * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS) | |
2234 | * to reduce average number of cache lines per packet. | |
2235 | * get_rps_cpus() for example only access one 64 bytes aligned block : | |
18e8c134 | 2236 | * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8) |
025be81e AB |
2237 | */ |
2238 | #ifndef NET_SKB_PAD | |
5933dd2f | 2239 | #define NET_SKB_PAD max(32, L1_CACHE_BYTES) |
025be81e AB |
2240 | #endif |
2241 | ||
7965bd4d | 2242 | int ___pskb_trim(struct sk_buff *skb, unsigned int len); |
1da177e4 LT |
2243 | |
2244 | static inline void __skb_trim(struct sk_buff *skb, unsigned int len) | |
2245 | { | |
c4264f27 | 2246 | if (unlikely(skb_is_nonlinear(skb))) { |
3cc0e873 HX |
2247 | WARN_ON(1); |
2248 | return; | |
2249 | } | |
27a884dc ACM |
2250 | skb->len = len; |
2251 | skb_set_tail_pointer(skb, len); | |
1da177e4 LT |
2252 | } |
2253 | ||
7965bd4d | 2254 | void skb_trim(struct sk_buff *skb, unsigned int len); |
1da177e4 LT |
2255 | |
2256 | static inline int __pskb_trim(struct sk_buff *skb, unsigned int len) | |
2257 | { | |
3cc0e873 HX |
2258 | if (skb->data_len) |
2259 | return ___pskb_trim(skb, len); | |
2260 | __skb_trim(skb, len); | |
2261 | return 0; | |
1da177e4 LT |
2262 | } |
2263 | ||
2264 | static inline int pskb_trim(struct sk_buff *skb, unsigned int len) | |
2265 | { | |
2266 | return (len < skb->len) ? __pskb_trim(skb, len) : 0; | |
2267 | } | |
2268 | ||
e9fa4f7b HX |
2269 | /** |
2270 | * pskb_trim_unique - remove end from a paged unique (not cloned) buffer | |
2271 | * @skb: buffer to alter | |
2272 | * @len: new length | |
2273 | * | |
2274 | * This is identical to pskb_trim except that the caller knows that | |
2275 | * the skb is not cloned so we should never get an error due to out- | |
2276 | * of-memory. | |
2277 | */ | |
2278 | static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len) | |
2279 | { | |
2280 | int err = pskb_trim(skb, len); | |
2281 | BUG_ON(err); | |
2282 | } | |
2283 | ||
1da177e4 LT |
2284 | /** |
2285 | * skb_orphan - orphan a buffer | |
2286 | * @skb: buffer to orphan | |
2287 | * | |
2288 | * If a buffer currently has an owner then we call the owner's | |
2289 | * destructor function and make the @skb unowned. The buffer continues | |
2290 | * to exist but is no longer charged to its former owner. | |
2291 | */ | |
2292 | static inline void skb_orphan(struct sk_buff *skb) | |
2293 | { | |
c34a7612 | 2294 | if (skb->destructor) { |
1da177e4 | 2295 | skb->destructor(skb); |
c34a7612 ED |
2296 | skb->destructor = NULL; |
2297 | skb->sk = NULL; | |
376c7311 ED |
2298 | } else { |
2299 | BUG_ON(skb->sk); | |
c34a7612 | 2300 | } |
1da177e4 LT |
2301 | } |
2302 | ||
a353e0ce MT |
2303 | /** |
2304 | * skb_orphan_frags - orphan the frags contained in a buffer | |
2305 | * @skb: buffer to orphan frags from | |
2306 | * @gfp_mask: allocation mask for replacement pages | |
2307 | * | |
2308 | * For each frag in the SKB which needs a destructor (i.e. has an | |
2309 | * owner) create a copy of that frag and release the original | |
2310 | * page by calling the destructor. | |
2311 | */ | |
2312 | static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask) | |
2313 | { | |
2314 | if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY))) | |
2315 | return 0; | |
2316 | return skb_copy_ubufs(skb, gfp_mask); | |
2317 | } | |
2318 | ||
1da177e4 LT |
2319 | /** |
2320 | * __skb_queue_purge - empty a list | |
2321 | * @list: list to empty | |
2322 | * | |
2323 | * Delete all buffers on an &sk_buff list. Each buffer is removed from | |
2324 | * the list and one reference dropped. This function does not take the | |
2325 | * list lock and the caller must hold the relevant locks to use it. | |
2326 | */ | |
7965bd4d | 2327 | void skb_queue_purge(struct sk_buff_head *list); |
1da177e4 LT |
2328 | static inline void __skb_queue_purge(struct sk_buff_head *list) |
2329 | { | |
2330 | struct sk_buff *skb; | |
2331 | while ((skb = __skb_dequeue(list)) != NULL) | |
2332 | kfree_skb(skb); | |
2333 | } | |
2334 | ||
7965bd4d | 2335 | void *netdev_alloc_frag(unsigned int fragsz); |
1da177e4 | 2336 | |
7965bd4d JP |
2337 | struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length, |
2338 | gfp_t gfp_mask); | |
8af27456 CH |
2339 | |
2340 | /** | |
2341 | * netdev_alloc_skb - allocate an skbuff for rx on a specific device | |
2342 | * @dev: network device to receive on | |
2343 | * @length: length to allocate | |
2344 | * | |
2345 | * Allocate a new &sk_buff and assign it a usage count of one. The | |
2346 | * buffer has unspecified headroom built in. Users should allocate | |
2347 | * the headroom they think they need without accounting for the | |
2348 | * built in space. The built in space is used for optimisations. | |
2349 | * | |
2350 | * %NULL is returned if there is no free memory. Although this function | |
2351 | * allocates memory it can be called from an interrupt. | |
2352 | */ | |
2353 | static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev, | |
6f532612 | 2354 | unsigned int length) |
8af27456 CH |
2355 | { |
2356 | return __netdev_alloc_skb(dev, length, GFP_ATOMIC); | |
2357 | } | |
2358 | ||
6f532612 ED |
2359 | /* legacy helper around __netdev_alloc_skb() */ |
2360 | static inline struct sk_buff *__dev_alloc_skb(unsigned int length, | |
2361 | gfp_t gfp_mask) | |
2362 | { | |
2363 | return __netdev_alloc_skb(NULL, length, gfp_mask); | |
2364 | } | |
2365 | ||
2366 | /* legacy helper around netdev_alloc_skb() */ | |
2367 | static inline struct sk_buff *dev_alloc_skb(unsigned int length) | |
2368 | { | |
2369 | return netdev_alloc_skb(NULL, length); | |
2370 | } | |
2371 | ||
2372 | ||
4915a0de ED |
2373 | static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev, |
2374 | unsigned int length, gfp_t gfp) | |
61321bbd | 2375 | { |
4915a0de | 2376 | struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp); |
61321bbd ED |
2377 | |
2378 | if (NET_IP_ALIGN && skb) | |
2379 | skb_reserve(skb, NET_IP_ALIGN); | |
2380 | return skb; | |
2381 | } | |
2382 | ||
4915a0de ED |
2383 | static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev, |
2384 | unsigned int length) | |
2385 | { | |
2386 | return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC); | |
2387 | } | |
2388 | ||
181edb2b AD |
2389 | static inline void skb_free_frag(void *addr) |
2390 | { | |
2391 | __free_page_frag(addr); | |
2392 | } | |
2393 | ||
ffde7328 | 2394 | void *napi_alloc_frag(unsigned int fragsz); |
fd11a83d AD |
2395 | struct sk_buff *__napi_alloc_skb(struct napi_struct *napi, |
2396 | unsigned int length, gfp_t gfp_mask); | |
2397 | static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi, | |
2398 | unsigned int length) | |
2399 | { | |
2400 | return __napi_alloc_skb(napi, length, GFP_ATOMIC); | |
2401 | } | |
ffde7328 | 2402 | |
71dfda58 AD |
2403 | /** |
2404 | * __dev_alloc_pages - allocate page for network Rx | |
2405 | * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx | |
2406 | * @order: size of the allocation | |
2407 | * | |
2408 | * Allocate a new page. | |
2409 | * | |
2410 | * %NULL is returned if there is no free memory. | |
2411 | */ | |
2412 | static inline struct page *__dev_alloc_pages(gfp_t gfp_mask, | |
2413 | unsigned int order) | |
2414 | { | |
2415 | /* This piece of code contains several assumptions. | |
2416 | * 1. This is for device Rx, therefor a cold page is preferred. | |
2417 | * 2. The expectation is the user wants a compound page. | |
2418 | * 3. If requesting a order 0 page it will not be compound | |
2419 | * due to the check to see if order has a value in prep_new_page | |
2420 | * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to | |
2421 | * code in gfp_to_alloc_flags that should be enforcing this. | |
2422 | */ | |
2423 | gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC; | |
2424 | ||
2425 | return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order); | |
2426 | } | |
2427 | ||
2428 | static inline struct page *dev_alloc_pages(unsigned int order) | |
2429 | { | |
2430 | return __dev_alloc_pages(GFP_ATOMIC, order); | |
2431 | } | |
2432 | ||
2433 | /** | |
2434 | * __dev_alloc_page - allocate a page for network Rx | |
2435 | * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx | |
2436 | * | |
2437 | * Allocate a new page. | |
2438 | * | |
2439 | * %NULL is returned if there is no free memory. | |
2440 | */ | |
2441 | static inline struct page *__dev_alloc_page(gfp_t gfp_mask) | |
2442 | { | |
2443 | return __dev_alloc_pages(gfp_mask, 0); | |
2444 | } | |
2445 | ||
2446 | static inline struct page *dev_alloc_page(void) | |
2447 | { | |
2448 | return __dev_alloc_page(GFP_ATOMIC); | |
2449 | } | |
2450 | ||
0614002b MG |
2451 | /** |
2452 | * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page | |
2453 | * @page: The page that was allocated from skb_alloc_page | |
2454 | * @skb: The skb that may need pfmemalloc set | |
2455 | */ | |
2456 | static inline void skb_propagate_pfmemalloc(struct page *page, | |
2457 | struct sk_buff *skb) | |
2458 | { | |
2f064f34 | 2459 | if (page_is_pfmemalloc(page)) |
0614002b MG |
2460 | skb->pfmemalloc = true; |
2461 | } | |
2462 | ||
131ea667 | 2463 | /** |
e227867f | 2464 | * skb_frag_page - retrieve the page referred to by a paged fragment |
131ea667 IC |
2465 | * @frag: the paged fragment |
2466 | * | |
2467 | * Returns the &struct page associated with @frag. | |
2468 | */ | |
2469 | static inline struct page *skb_frag_page(const skb_frag_t *frag) | |
2470 | { | |
a8605c60 | 2471 | return frag->page.p; |
131ea667 IC |
2472 | } |
2473 | ||
2474 | /** | |
2475 | * __skb_frag_ref - take an addition reference on a paged fragment. | |
2476 | * @frag: the paged fragment | |
2477 | * | |
2478 | * Takes an additional reference on the paged fragment @frag. | |
2479 | */ | |
2480 | static inline void __skb_frag_ref(skb_frag_t *frag) | |
2481 | { | |
2482 | get_page(skb_frag_page(frag)); | |
2483 | } | |
2484 | ||
2485 | /** | |
2486 | * skb_frag_ref - take an addition reference on a paged fragment of an skb. | |
2487 | * @skb: the buffer | |
2488 | * @f: the fragment offset. | |
2489 | * | |
2490 | * Takes an additional reference on the @f'th paged fragment of @skb. | |
2491 | */ | |
2492 | static inline void skb_frag_ref(struct sk_buff *skb, int f) | |
2493 | { | |
2494 | __skb_frag_ref(&skb_shinfo(skb)->frags[f]); | |
2495 | } | |
2496 | ||
2497 | /** | |
2498 | * __skb_frag_unref - release a reference on a paged fragment. | |
2499 | * @frag: the paged fragment | |
2500 | * | |
2501 | * Releases a reference on the paged fragment @frag. | |
2502 | */ | |
2503 | static inline void __skb_frag_unref(skb_frag_t *frag) | |
2504 | { | |
2505 | put_page(skb_frag_page(frag)); | |
2506 | } | |
2507 | ||
2508 | /** | |
2509 | * skb_frag_unref - release a reference on a paged fragment of an skb. | |
2510 | * @skb: the buffer | |
2511 | * @f: the fragment offset | |
2512 | * | |
2513 | * Releases a reference on the @f'th paged fragment of @skb. | |
2514 | */ | |
2515 | static inline void skb_frag_unref(struct sk_buff *skb, int f) | |
2516 | { | |
2517 | __skb_frag_unref(&skb_shinfo(skb)->frags[f]); | |
2518 | } | |
2519 | ||
2520 | /** | |
2521 | * skb_frag_address - gets the address of the data contained in a paged fragment | |
2522 | * @frag: the paged fragment buffer | |
2523 | * | |
2524 | * Returns the address of the data within @frag. The page must already | |
2525 | * be mapped. | |
2526 | */ | |
2527 | static inline void *skb_frag_address(const skb_frag_t *frag) | |
2528 | { | |
2529 | return page_address(skb_frag_page(frag)) + frag->page_offset; | |
2530 | } | |
2531 | ||
2532 | /** | |
2533 | * skb_frag_address_safe - gets the address of the data contained in a paged fragment | |
2534 | * @frag: the paged fragment buffer | |
2535 | * | |
2536 | * Returns the address of the data within @frag. Checks that the page | |
2537 | * is mapped and returns %NULL otherwise. | |
2538 | */ | |
2539 | static inline void *skb_frag_address_safe(const skb_frag_t *frag) | |
2540 | { | |
2541 | void *ptr = page_address(skb_frag_page(frag)); | |
2542 | if (unlikely(!ptr)) | |
2543 | return NULL; | |
2544 | ||
2545 | return ptr + frag->page_offset; | |
2546 | } | |
2547 | ||
2548 | /** | |
2549 | * __skb_frag_set_page - sets the page contained in a paged fragment | |
2550 | * @frag: the paged fragment | |
2551 | * @page: the page to set | |
2552 | * | |
2553 | * Sets the fragment @frag to contain @page. | |
2554 | */ | |
2555 | static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page) | |
2556 | { | |
a8605c60 | 2557 | frag->page.p = page; |
131ea667 IC |
2558 | } |
2559 | ||
2560 | /** | |
2561 | * skb_frag_set_page - sets the page contained in a paged fragment of an skb | |
2562 | * @skb: the buffer | |
2563 | * @f: the fragment offset | |
2564 | * @page: the page to set | |
2565 | * | |
2566 | * Sets the @f'th fragment of @skb to contain @page. | |
2567 | */ | |
2568 | static inline void skb_frag_set_page(struct sk_buff *skb, int f, | |
2569 | struct page *page) | |
2570 | { | |
2571 | __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page); | |
2572 | } | |
2573 | ||
400dfd3a ED |
2574 | bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio); |
2575 | ||
131ea667 IC |
2576 | /** |
2577 | * skb_frag_dma_map - maps a paged fragment via the DMA API | |
f83347df | 2578 | * @dev: the device to map the fragment to |
131ea667 IC |
2579 | * @frag: the paged fragment to map |
2580 | * @offset: the offset within the fragment (starting at the | |
2581 | * fragment's own offset) | |
2582 | * @size: the number of bytes to map | |
f83347df | 2583 | * @dir: the direction of the mapping (%PCI_DMA_*) |
131ea667 IC |
2584 | * |
2585 | * Maps the page associated with @frag to @device. | |
2586 | */ | |
2587 | static inline dma_addr_t skb_frag_dma_map(struct device *dev, | |
2588 | const skb_frag_t *frag, | |
2589 | size_t offset, size_t size, | |
2590 | enum dma_data_direction dir) | |
2591 | { | |
2592 | return dma_map_page(dev, skb_frag_page(frag), | |
2593 | frag->page_offset + offset, size, dir); | |
2594 | } | |
2595 | ||
117632e6 ED |
2596 | static inline struct sk_buff *pskb_copy(struct sk_buff *skb, |
2597 | gfp_t gfp_mask) | |
2598 | { | |
2599 | return __pskb_copy(skb, skb_headroom(skb), gfp_mask); | |
2600 | } | |
2601 | ||
bad93e9d OP |
2602 | |
2603 | static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb, | |
2604 | gfp_t gfp_mask) | |
2605 | { | |
2606 | return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true); | |
2607 | } | |
2608 | ||
2609 | ||
334a8132 PM |
2610 | /** |
2611 | * skb_clone_writable - is the header of a clone writable | |
2612 | * @skb: buffer to check | |
2613 | * @len: length up to which to write | |
2614 | * | |
2615 | * Returns true if modifying the header part of the cloned buffer | |
2616 | * does not requires the data to be copied. | |
2617 | */ | |
05bdd2f1 | 2618 | static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len) |
334a8132 PM |
2619 | { |
2620 | return !skb_header_cloned(skb) && | |
2621 | skb_headroom(skb) + len <= skb->hdr_len; | |
2622 | } | |
2623 | ||
d9cc2048 HX |
2624 | static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom, |
2625 | int cloned) | |
2626 | { | |
2627 | int delta = 0; | |
2628 | ||
d9cc2048 HX |
2629 | if (headroom > skb_headroom(skb)) |
2630 | delta = headroom - skb_headroom(skb); | |
2631 | ||
2632 | if (delta || cloned) | |
2633 | return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0, | |
2634 | GFP_ATOMIC); | |
2635 | return 0; | |
2636 | } | |
2637 | ||
1da177e4 LT |
2638 | /** |
2639 | * skb_cow - copy header of skb when it is required | |
2640 | * @skb: buffer to cow | |
2641 | * @headroom: needed headroom | |
2642 | * | |
2643 | * If the skb passed lacks sufficient headroom or its data part | |
2644 | * is shared, data is reallocated. If reallocation fails, an error | |
2645 | * is returned and original skb is not changed. | |
2646 | * | |
2647 | * The result is skb with writable area skb->head...skb->tail | |
2648 | * and at least @headroom of space at head. | |
2649 | */ | |
2650 | static inline int skb_cow(struct sk_buff *skb, unsigned int headroom) | |
2651 | { | |
d9cc2048 HX |
2652 | return __skb_cow(skb, headroom, skb_cloned(skb)); |
2653 | } | |
1da177e4 | 2654 | |
d9cc2048 HX |
2655 | /** |
2656 | * skb_cow_head - skb_cow but only making the head writable | |
2657 | * @skb: buffer to cow | |
2658 | * @headroom: needed headroom | |
2659 | * | |
2660 | * This function is identical to skb_cow except that we replace the | |
2661 | * skb_cloned check by skb_header_cloned. It should be used when | |
2662 | * you only need to push on some header and do not need to modify | |
2663 | * the data. | |
2664 | */ | |
2665 | static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom) | |
2666 | { | |
2667 | return __skb_cow(skb, headroom, skb_header_cloned(skb)); | |
1da177e4 LT |
2668 | } |
2669 | ||
2670 | /** | |
2671 | * skb_padto - pad an skbuff up to a minimal size | |
2672 | * @skb: buffer to pad | |
2673 | * @len: minimal length | |
2674 | * | |
2675 | * Pads up a buffer to ensure the trailing bytes exist and are | |
2676 | * blanked. If the buffer already contains sufficient data it | |
5b057c6b HX |
2677 | * is untouched. Otherwise it is extended. Returns zero on |
2678 | * success. The skb is freed on error. | |
1da177e4 | 2679 | */ |
5b057c6b | 2680 | static inline int skb_padto(struct sk_buff *skb, unsigned int len) |
1da177e4 LT |
2681 | { |
2682 | unsigned int size = skb->len; | |
2683 | if (likely(size >= len)) | |
5b057c6b | 2684 | return 0; |
987c402a | 2685 | return skb_pad(skb, len - size); |
1da177e4 LT |
2686 | } |
2687 | ||
9c0c1124 AD |
2688 | /** |
2689 | * skb_put_padto - increase size and pad an skbuff up to a minimal size | |
2690 | * @skb: buffer to pad | |
2691 | * @len: minimal length | |
2692 | * | |
2693 | * Pads up a buffer to ensure the trailing bytes exist and are | |
2694 | * blanked. If the buffer already contains sufficient data it | |
2695 | * is untouched. Otherwise it is extended. Returns zero on | |
2696 | * success. The skb is freed on error. | |
2697 | */ | |
2698 | static inline int skb_put_padto(struct sk_buff *skb, unsigned int len) | |
2699 | { | |
2700 | unsigned int size = skb->len; | |
2701 | ||
2702 | if (unlikely(size < len)) { | |
2703 | len -= size; | |
2704 | if (skb_pad(skb, len)) | |
2705 | return -ENOMEM; | |
2706 | __skb_put(skb, len); | |
2707 | } | |
2708 | return 0; | |
2709 | } | |
2710 | ||
1da177e4 | 2711 | static inline int skb_add_data(struct sk_buff *skb, |
af2b040e | 2712 | struct iov_iter *from, int copy) |
1da177e4 LT |
2713 | { |
2714 | const int off = skb->len; | |
2715 | ||
2716 | if (skb->ip_summed == CHECKSUM_NONE) { | |
af2b040e AV |
2717 | __wsum csum = 0; |
2718 | if (csum_and_copy_from_iter(skb_put(skb, copy), copy, | |
2719 | &csum, from) == copy) { | |
1da177e4 LT |
2720 | skb->csum = csum_block_add(skb->csum, csum, off); |
2721 | return 0; | |
2722 | } | |
af2b040e | 2723 | } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy) |
1da177e4 LT |
2724 | return 0; |
2725 | ||
2726 | __skb_trim(skb, off); | |
2727 | return -EFAULT; | |
2728 | } | |
2729 | ||
38ba0a65 ED |
2730 | static inline bool skb_can_coalesce(struct sk_buff *skb, int i, |
2731 | const struct page *page, int off) | |
1da177e4 LT |
2732 | { |
2733 | if (i) { | |
9e903e08 | 2734 | const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1]; |
1da177e4 | 2735 | |
ea2ab693 | 2736 | return page == skb_frag_page(frag) && |
9e903e08 | 2737 | off == frag->page_offset + skb_frag_size(frag); |
1da177e4 | 2738 | } |
38ba0a65 | 2739 | return false; |
1da177e4 LT |
2740 | } |
2741 | ||
364c6bad HX |
2742 | static inline int __skb_linearize(struct sk_buff *skb) |
2743 | { | |
2744 | return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM; | |
2745 | } | |
2746 | ||
1da177e4 LT |
2747 | /** |
2748 | * skb_linearize - convert paged skb to linear one | |
2749 | * @skb: buffer to linarize | |
1da177e4 LT |
2750 | * |
2751 | * If there is no free memory -ENOMEM is returned, otherwise zero | |
2752 | * is returned and the old skb data released. | |
2753 | */ | |
364c6bad HX |
2754 | static inline int skb_linearize(struct sk_buff *skb) |
2755 | { | |
2756 | return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0; | |
2757 | } | |
2758 | ||
cef401de ED |
2759 | /** |
2760 | * skb_has_shared_frag - can any frag be overwritten | |
2761 | * @skb: buffer to test | |
2762 | * | |
2763 | * Return true if the skb has at least one frag that might be modified | |
2764 | * by an external entity (as in vmsplice()/sendfile()) | |
2765 | */ | |
2766 | static inline bool skb_has_shared_frag(const struct sk_buff *skb) | |
2767 | { | |
c9af6db4 PS |
2768 | return skb_is_nonlinear(skb) && |
2769 | skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG; | |
cef401de ED |
2770 | } |
2771 | ||
364c6bad HX |
2772 | /** |
2773 | * skb_linearize_cow - make sure skb is linear and writable | |
2774 | * @skb: buffer to process | |
2775 | * | |
2776 | * If there is no free memory -ENOMEM is returned, otherwise zero | |
2777 | * is returned and the old skb data released. | |
2778 | */ | |
2779 | static inline int skb_linearize_cow(struct sk_buff *skb) | |
1da177e4 | 2780 | { |
364c6bad HX |
2781 | return skb_is_nonlinear(skb) || skb_cloned(skb) ? |
2782 | __skb_linearize(skb) : 0; | |
1da177e4 LT |
2783 | } |
2784 | ||
2785 | /** | |
2786 | * skb_postpull_rcsum - update checksum for received skb after pull | |
2787 | * @skb: buffer to update | |
2788 | * @start: start of data before pull | |
2789 | * @len: length of data pulled | |
2790 | * | |
2791 | * After doing a pull on a received packet, you need to call this to | |
84fa7933 PM |
2792 | * update the CHECKSUM_COMPLETE checksum, or set ip_summed to |
2793 | * CHECKSUM_NONE so that it can be recomputed from scratch. | |
1da177e4 LT |
2794 | */ |
2795 | ||
2796 | static inline void skb_postpull_rcsum(struct sk_buff *skb, | |
cbb042f9 | 2797 | const void *start, unsigned int len) |
1da177e4 | 2798 | { |
84fa7933 | 2799 | if (skb->ip_summed == CHECKSUM_COMPLETE) |
1da177e4 | 2800 | skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0)); |
6ae459bd | 2801 | else if (skb->ip_summed == CHECKSUM_PARTIAL && |
31b33dfb | 2802 | skb_checksum_start_offset(skb) < 0) |
6ae459bd | 2803 | skb->ip_summed = CHECKSUM_NONE; |
1da177e4 LT |
2804 | } |
2805 | ||
cbb042f9 HX |
2806 | unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len); |
2807 | ||
f8ffad69 DB |
2808 | static inline void skb_postpush_rcsum(struct sk_buff *skb, |
2809 | const void *start, unsigned int len) | |
2810 | { | |
2811 | /* For performing the reverse operation to skb_postpull_rcsum(), | |
2812 | * we can instead of ... | |
2813 | * | |
2814 | * skb->csum = csum_add(skb->csum, csum_partial(start, len, 0)); | |
2815 | * | |
2816 | * ... just use this equivalent version here to save a few | |
2817 | * instructions. Feeding csum of 0 in csum_partial() and later | |
2818 | * on adding skb->csum is equivalent to feed skb->csum in the | |
2819 | * first place. | |
2820 | */ | |
2821 | if (skb->ip_summed == CHECKSUM_COMPLETE) | |
2822 | skb->csum = csum_partial(start, len, skb->csum); | |
2823 | } | |
2824 | ||
7ce5a27f DM |
2825 | /** |
2826 | * pskb_trim_rcsum - trim received skb and update checksum | |
2827 | * @skb: buffer to trim | |
2828 | * @len: new length | |
2829 | * | |
2830 | * This is exactly the same as pskb_trim except that it ensures the | |
2831 | * checksum of received packets are still valid after the operation. | |
2832 | */ | |
2833 | ||
2834 | static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len) | |
2835 | { | |
2836 | if (likely(len >= skb->len)) | |
2837 | return 0; | |
2838 | if (skb->ip_summed == CHECKSUM_COMPLETE) | |
2839 | skb->ip_summed = CHECKSUM_NONE; | |
2840 | return __pskb_trim(skb, len); | |
2841 | } | |
2842 | ||
1da177e4 LT |
2843 | #define skb_queue_walk(queue, skb) \ |
2844 | for (skb = (queue)->next; \ | |
a1e4891f | 2845 | skb != (struct sk_buff *)(queue); \ |
1da177e4 LT |
2846 | skb = skb->next) |
2847 | ||
46f8914e JC |
2848 | #define skb_queue_walk_safe(queue, skb, tmp) \ |
2849 | for (skb = (queue)->next, tmp = skb->next; \ | |
2850 | skb != (struct sk_buff *)(queue); \ | |
2851 | skb = tmp, tmp = skb->next) | |
2852 | ||
1164f52a | 2853 | #define skb_queue_walk_from(queue, skb) \ |
a1e4891f | 2854 | for (; skb != (struct sk_buff *)(queue); \ |
1164f52a DM |
2855 | skb = skb->next) |
2856 | ||
2857 | #define skb_queue_walk_from_safe(queue, skb, tmp) \ | |
2858 | for (tmp = skb->next; \ | |
2859 | skb != (struct sk_buff *)(queue); \ | |
2860 | skb = tmp, tmp = skb->next) | |
2861 | ||
300ce174 SH |
2862 | #define skb_queue_reverse_walk(queue, skb) \ |
2863 | for (skb = (queue)->prev; \ | |
a1e4891f | 2864 | skb != (struct sk_buff *)(queue); \ |
300ce174 SH |
2865 | skb = skb->prev) |
2866 | ||
686a2955 DM |
2867 | #define skb_queue_reverse_walk_safe(queue, skb, tmp) \ |
2868 | for (skb = (queue)->prev, tmp = skb->prev; \ | |
2869 | skb != (struct sk_buff *)(queue); \ | |
2870 | skb = tmp, tmp = skb->prev) | |
2871 | ||
2872 | #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \ | |
2873 | for (tmp = skb->prev; \ | |
2874 | skb != (struct sk_buff *)(queue); \ | |
2875 | skb = tmp, tmp = skb->prev) | |
1da177e4 | 2876 | |
21dc3301 | 2877 | static inline bool skb_has_frag_list(const struct sk_buff *skb) |
ee039871 DM |
2878 | { |
2879 | return skb_shinfo(skb)->frag_list != NULL; | |
2880 | } | |
2881 | ||
2882 | static inline void skb_frag_list_init(struct sk_buff *skb) | |
2883 | { | |
2884 | skb_shinfo(skb)->frag_list = NULL; | |
2885 | } | |
2886 | ||
ee039871 DM |
2887 | #define skb_walk_frags(skb, iter) \ |
2888 | for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next) | |
2889 | ||
ea3793ee RW |
2890 | |
2891 | int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p, | |
2892 | const struct sk_buff *skb); | |
2893 | struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags, | |
2894 | int *peeked, int *off, int *err, | |
2895 | struct sk_buff **last); | |
7965bd4d JP |
2896 | struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags, |
2897 | int *peeked, int *off, int *err); | |
2898 | struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock, | |
2899 | int *err); | |
2900 | unsigned int datagram_poll(struct file *file, struct socket *sock, | |
2901 | struct poll_table_struct *wait); | |
c0371da6 AV |
2902 | int skb_copy_datagram_iter(const struct sk_buff *from, int offset, |
2903 | struct iov_iter *to, int size); | |
51f3d02b DM |
2904 | static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset, |
2905 | struct msghdr *msg, int size) | |
2906 | { | |
e5a4b0bb | 2907 | return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size); |
227158db | 2908 | } |
e5a4b0bb AV |
2909 | int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen, |
2910 | struct msghdr *msg); | |
3a654f97 AV |
2911 | int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset, |
2912 | struct iov_iter *from, int len); | |
3a654f97 | 2913 | int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm); |
7965bd4d JP |
2914 | void skb_free_datagram(struct sock *sk, struct sk_buff *skb); |
2915 | void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb); | |
2916 | int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags); | |
7965bd4d JP |
2917 | int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len); |
2918 | int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len); | |
2919 | __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to, | |
2920 | int len, __wsum csum); | |
a60e3cc7 HFS |
2921 | ssize_t skb_socket_splice(struct sock *sk, |
2922 | struct pipe_inode_info *pipe, | |
2923 | struct splice_pipe_desc *spd); | |
2924 | int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset, | |
7965bd4d | 2925 | struct pipe_inode_info *pipe, unsigned int len, |
a60e3cc7 HFS |
2926 | unsigned int flags, |
2927 | ssize_t (*splice_cb)(struct sock *, | |
2928 | struct pipe_inode_info *, | |
2929 | struct splice_pipe_desc *)); | |
7965bd4d | 2930 | void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to); |
af2806f8 | 2931 | unsigned int skb_zerocopy_headlen(const struct sk_buff *from); |
36d5fe6a ZK |
2932 | int skb_zerocopy(struct sk_buff *to, struct sk_buff *from, |
2933 | int len, int hlen); | |
7965bd4d JP |
2934 | void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len); |
2935 | int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen); | |
2936 | void skb_scrub_packet(struct sk_buff *skb, bool xnet); | |
de960aa9 | 2937 | unsigned int skb_gso_transport_seglen(const struct sk_buff *skb); |
7965bd4d | 2938 | struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features); |
0d5501c1 | 2939 | struct sk_buff *skb_vlan_untag(struct sk_buff *skb); |
e2195121 | 2940 | int skb_ensure_writable(struct sk_buff *skb, int write_len); |
93515d53 JP |
2941 | int skb_vlan_pop(struct sk_buff *skb); |
2942 | int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci); | |
20380731 | 2943 | |
6ce8e9ce AV |
2944 | static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len) |
2945 | { | |
21226abb | 2946 | return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT; |
6ce8e9ce AV |
2947 | } |
2948 | ||
7eab8d9e AV |
2949 | static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len) |
2950 | { | |
e5a4b0bb | 2951 | return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT; |
7eab8d9e AV |
2952 | } |
2953 | ||
2817a336 DB |
2954 | struct skb_checksum_ops { |
2955 | __wsum (*update)(const void *mem, int len, __wsum wsum); | |
2956 | __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len); | |
2957 | }; | |
2958 | ||
2959 | __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len, | |
2960 | __wsum csum, const struct skb_checksum_ops *ops); | |
2961 | __wsum skb_checksum(const struct sk_buff *skb, int offset, int len, | |
2962 | __wsum csum); | |
2963 | ||
1e98a0f0 ED |
2964 | static inline void * __must_check |
2965 | __skb_header_pointer(const struct sk_buff *skb, int offset, | |
2966 | int len, void *data, int hlen, void *buffer) | |
1da177e4 | 2967 | { |
55820ee2 | 2968 | if (hlen - offset >= len) |
690e36e7 | 2969 | return data + offset; |
1da177e4 | 2970 | |
690e36e7 DM |
2971 | if (!skb || |
2972 | skb_copy_bits(skb, offset, buffer, len) < 0) | |
1da177e4 LT |
2973 | return NULL; |
2974 | ||
2975 | return buffer; | |
2976 | } | |
2977 | ||
1e98a0f0 ED |
2978 | static inline void * __must_check |
2979 | skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer) | |
690e36e7 DM |
2980 | { |
2981 | return __skb_header_pointer(skb, offset, len, skb->data, | |
2982 | skb_headlen(skb), buffer); | |
2983 | } | |
2984 | ||
4262e5cc DB |
2985 | /** |
2986 | * skb_needs_linearize - check if we need to linearize a given skb | |
2987 | * depending on the given device features. | |
2988 | * @skb: socket buffer to check | |
2989 | * @features: net device features | |
2990 | * | |
2991 | * Returns true if either: | |
2992 | * 1. skb has frag_list and the device doesn't support FRAGLIST, or | |
2993 | * 2. skb is fragmented and the device does not support SG. | |
2994 | */ | |
2995 | static inline bool skb_needs_linearize(struct sk_buff *skb, | |
2996 | netdev_features_t features) | |
2997 | { | |
2998 | return skb_is_nonlinear(skb) && | |
2999 | ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) || | |
3000 | (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG))); | |
3001 | } | |
3002 | ||
d626f62b ACM |
3003 | static inline void skb_copy_from_linear_data(const struct sk_buff *skb, |
3004 | void *to, | |
3005 | const unsigned int len) | |
3006 | { | |
3007 | memcpy(to, skb->data, len); | |
3008 | } | |
3009 | ||
3010 | static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb, | |
3011 | const int offset, void *to, | |
3012 | const unsigned int len) | |
3013 | { | |
3014 | memcpy(to, skb->data + offset, len); | |
3015 | } | |
3016 | ||
27d7ff46 ACM |
3017 | static inline void skb_copy_to_linear_data(struct sk_buff *skb, |
3018 | const void *from, | |
3019 | const unsigned int len) | |
3020 | { | |
3021 | memcpy(skb->data, from, len); | |
3022 | } | |
3023 | ||
3024 | static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb, | |
3025 | const int offset, | |
3026 | const void *from, | |
3027 | const unsigned int len) | |
3028 | { | |
3029 | memcpy(skb->data + offset, from, len); | |
3030 | } | |
3031 | ||
7965bd4d | 3032 | void skb_init(void); |
1da177e4 | 3033 | |
ac45f602 PO |
3034 | static inline ktime_t skb_get_ktime(const struct sk_buff *skb) |
3035 | { | |
3036 | return skb->tstamp; | |
3037 | } | |
3038 | ||
a61bbcf2 PM |
3039 | /** |
3040 | * skb_get_timestamp - get timestamp from a skb | |
3041 | * @skb: skb to get stamp from | |
3042 | * @stamp: pointer to struct timeval to store stamp in | |
3043 | * | |
3044 | * Timestamps are stored in the skb as offsets to a base timestamp. | |
3045 | * This function converts the offset back to a struct timeval and stores | |
3046 | * it in stamp. | |
3047 | */ | |
ac45f602 PO |
3048 | static inline void skb_get_timestamp(const struct sk_buff *skb, |
3049 | struct timeval *stamp) | |
a61bbcf2 | 3050 | { |
b7aa0bf7 | 3051 | *stamp = ktime_to_timeval(skb->tstamp); |
a61bbcf2 PM |
3052 | } |
3053 | ||
ac45f602 PO |
3054 | static inline void skb_get_timestampns(const struct sk_buff *skb, |
3055 | struct timespec *stamp) | |
3056 | { | |
3057 | *stamp = ktime_to_timespec(skb->tstamp); | |
3058 | } | |
3059 | ||
b7aa0bf7 | 3060 | static inline void __net_timestamp(struct sk_buff *skb) |
a61bbcf2 | 3061 | { |
b7aa0bf7 | 3062 | skb->tstamp = ktime_get_real(); |
a61bbcf2 PM |
3063 | } |
3064 | ||
164891aa SH |
3065 | static inline ktime_t net_timedelta(ktime_t t) |
3066 | { | |
3067 | return ktime_sub(ktime_get_real(), t); | |
3068 | } | |
3069 | ||
b9ce204f IJ |
3070 | static inline ktime_t net_invalid_timestamp(void) |
3071 | { | |
3072 | return ktime_set(0, 0); | |
3073 | } | |
a61bbcf2 | 3074 | |
62bccb8c AD |
3075 | struct sk_buff *skb_clone_sk(struct sk_buff *skb); |
3076 | ||
c1f19b51 RC |
3077 | #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING |
3078 | ||
7965bd4d JP |
3079 | void skb_clone_tx_timestamp(struct sk_buff *skb); |
3080 | bool skb_defer_rx_timestamp(struct sk_buff *skb); | |
c1f19b51 RC |
3081 | |
3082 | #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */ | |
3083 | ||
3084 | static inline void skb_clone_tx_timestamp(struct sk_buff *skb) | |
3085 | { | |
3086 | } | |
3087 | ||
3088 | static inline bool skb_defer_rx_timestamp(struct sk_buff *skb) | |
3089 | { | |
3090 | return false; | |
3091 | } | |
3092 | ||
3093 | #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */ | |
3094 | ||
3095 | /** | |
3096 | * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps | |
3097 | * | |
da92b194 RC |
3098 | * PHY drivers may accept clones of transmitted packets for |
3099 | * timestamping via their phy_driver.txtstamp method. These drivers | |
7a76a021 BP |
3100 | * must call this function to return the skb back to the stack with a |
3101 | * timestamp. | |
da92b194 | 3102 | * |
c1f19b51 | 3103 | * @skb: clone of the the original outgoing packet |
7a76a021 | 3104 | * @hwtstamps: hardware time stamps |
c1f19b51 RC |
3105 | * |
3106 | */ | |
3107 | void skb_complete_tx_timestamp(struct sk_buff *skb, | |
3108 | struct skb_shared_hwtstamps *hwtstamps); | |
3109 | ||
e7fd2885 WB |
3110 | void __skb_tstamp_tx(struct sk_buff *orig_skb, |
3111 | struct skb_shared_hwtstamps *hwtstamps, | |
3112 | struct sock *sk, int tstype); | |
3113 | ||
ac45f602 PO |
3114 | /** |
3115 | * skb_tstamp_tx - queue clone of skb with send time stamps | |
3116 | * @orig_skb: the original outgoing packet | |
3117 | * @hwtstamps: hardware time stamps, may be NULL if not available | |
3118 | * | |
3119 | * If the skb has a socket associated, then this function clones the | |
3120 | * skb (thus sharing the actual data and optional structures), stores | |
3121 | * the optional hardware time stamping information (if non NULL) or | |
3122 | * generates a software time stamp (otherwise), then queues the clone | |
3123 | * to the error queue of the socket. Errors are silently ignored. | |
3124 | */ | |
7965bd4d JP |
3125 | void skb_tstamp_tx(struct sk_buff *orig_skb, |
3126 | struct skb_shared_hwtstamps *hwtstamps); | |
ac45f602 | 3127 | |
4507a715 RC |
3128 | static inline void sw_tx_timestamp(struct sk_buff *skb) |
3129 | { | |
2244d07b OH |
3130 | if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP && |
3131 | !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS)) | |
4507a715 RC |
3132 | skb_tstamp_tx(skb, NULL); |
3133 | } | |
3134 | ||
3135 | /** | |
3136 | * skb_tx_timestamp() - Driver hook for transmit timestamping | |
3137 | * | |
3138 | * Ethernet MAC Drivers should call this function in their hard_xmit() | |
4ff75b7c | 3139 | * function immediately before giving the sk_buff to the MAC hardware. |
4507a715 | 3140 | * |
73409f3b DM |
3141 | * Specifically, one should make absolutely sure that this function is |
3142 | * called before TX completion of this packet can trigger. Otherwise | |
3143 | * the packet could potentially already be freed. | |
3144 | * | |
4507a715 RC |
3145 | * @skb: A socket buffer. |
3146 | */ | |
3147 | static inline void skb_tx_timestamp(struct sk_buff *skb) | |
3148 | { | |
c1f19b51 | 3149 | skb_clone_tx_timestamp(skb); |
4507a715 RC |
3150 | sw_tx_timestamp(skb); |
3151 | } | |
3152 | ||
6e3e939f JB |
3153 | /** |
3154 | * skb_complete_wifi_ack - deliver skb with wifi status | |
3155 | * | |
3156 | * @skb: the original outgoing packet | |
3157 | * @acked: ack status | |
3158 | * | |
3159 | */ | |
3160 | void skb_complete_wifi_ack(struct sk_buff *skb, bool acked); | |
3161 | ||
7965bd4d JP |
3162 | __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len); |
3163 | __sum16 __skb_checksum_complete(struct sk_buff *skb); | |
fb286bb2 | 3164 | |
60476372 HX |
3165 | static inline int skb_csum_unnecessary(const struct sk_buff *skb) |
3166 | { | |
6edec0e6 TH |
3167 | return ((skb->ip_summed == CHECKSUM_UNNECESSARY) || |
3168 | skb->csum_valid || | |
3169 | (skb->ip_summed == CHECKSUM_PARTIAL && | |
3170 | skb_checksum_start_offset(skb) >= 0)); | |
60476372 HX |
3171 | } |
3172 | ||
fb286bb2 HX |
3173 | /** |
3174 | * skb_checksum_complete - Calculate checksum of an entire packet | |
3175 | * @skb: packet to process | |
3176 | * | |
3177 | * This function calculates the checksum over the entire packet plus | |
3178 | * the value of skb->csum. The latter can be used to supply the | |
3179 | * checksum of a pseudo header as used by TCP/UDP. It returns the | |
3180 | * checksum. | |
3181 | * | |
3182 | * For protocols that contain complete checksums such as ICMP/TCP/UDP, | |
3183 | * this function can be used to verify that checksum on received | |
3184 | * packets. In that case the function should return zero if the | |
3185 | * checksum is correct. In particular, this function will return zero | |
3186 | * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the | |
3187 | * hardware has already verified the correctness of the checksum. | |
3188 | */ | |
4381ca3c | 3189 | static inline __sum16 skb_checksum_complete(struct sk_buff *skb) |
fb286bb2 | 3190 | { |
60476372 HX |
3191 | return skb_csum_unnecessary(skb) ? |
3192 | 0 : __skb_checksum_complete(skb); | |
fb286bb2 HX |
3193 | } |
3194 | ||
77cffe23 TH |
3195 | static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb) |
3196 | { | |
3197 | if (skb->ip_summed == CHECKSUM_UNNECESSARY) { | |
3198 | if (skb->csum_level == 0) | |
3199 | skb->ip_summed = CHECKSUM_NONE; | |
3200 | else | |
3201 | skb->csum_level--; | |
3202 | } | |
3203 | } | |
3204 | ||
3205 | static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb) | |
3206 | { | |
3207 | if (skb->ip_summed == CHECKSUM_UNNECESSARY) { | |
3208 | if (skb->csum_level < SKB_MAX_CSUM_LEVEL) | |
3209 | skb->csum_level++; | |
3210 | } else if (skb->ip_summed == CHECKSUM_NONE) { | |
3211 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
3212 | skb->csum_level = 0; | |
3213 | } | |
3214 | } | |
3215 | ||
5a212329 TH |
3216 | static inline void __skb_mark_checksum_bad(struct sk_buff *skb) |
3217 | { | |
3218 | /* Mark current checksum as bad (typically called from GRO | |
3219 | * path). In the case that ip_summed is CHECKSUM_NONE | |
3220 | * this must be the first checksum encountered in the packet. | |
3221 | * When ip_summed is CHECKSUM_UNNECESSARY, this is the first | |
3222 | * checksum after the last one validated. For UDP, a zero | |
3223 | * checksum can not be marked as bad. | |
3224 | */ | |
3225 | ||
3226 | if (skb->ip_summed == CHECKSUM_NONE || | |
3227 | skb->ip_summed == CHECKSUM_UNNECESSARY) | |
3228 | skb->csum_bad = 1; | |
3229 | } | |
3230 | ||
76ba0aae TH |
3231 | /* Check if we need to perform checksum complete validation. |
3232 | * | |
3233 | * Returns true if checksum complete is needed, false otherwise | |
3234 | * (either checksum is unnecessary or zero checksum is allowed). | |
3235 | */ | |
3236 | static inline bool __skb_checksum_validate_needed(struct sk_buff *skb, | |
3237 | bool zero_okay, | |
3238 | __sum16 check) | |
3239 | { | |
5d0c2b95 TH |
3240 | if (skb_csum_unnecessary(skb) || (zero_okay && !check)) { |
3241 | skb->csum_valid = 1; | |
77cffe23 | 3242 | __skb_decr_checksum_unnecessary(skb); |
76ba0aae TH |
3243 | return false; |
3244 | } | |
3245 | ||
3246 | return true; | |
3247 | } | |
3248 | ||
3249 | /* For small packets <= CHECKSUM_BREAK peform checksum complete directly | |
3250 | * in checksum_init. | |
3251 | */ | |
3252 | #define CHECKSUM_BREAK 76 | |
3253 | ||
4e18b9ad TH |
3254 | /* Unset checksum-complete |
3255 | * | |
3256 | * Unset checksum complete can be done when packet is being modified | |
3257 | * (uncompressed for instance) and checksum-complete value is | |
3258 | * invalidated. | |
3259 | */ | |
3260 | static inline void skb_checksum_complete_unset(struct sk_buff *skb) | |
3261 | { | |
3262 | if (skb->ip_summed == CHECKSUM_COMPLETE) | |
3263 | skb->ip_summed = CHECKSUM_NONE; | |
3264 | } | |
3265 | ||
76ba0aae TH |
3266 | /* Validate (init) checksum based on checksum complete. |
3267 | * | |
3268 | * Return values: | |
3269 | * 0: checksum is validated or try to in skb_checksum_complete. In the latter | |
3270 | * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo | |
3271 | * checksum is stored in skb->csum for use in __skb_checksum_complete | |
3272 | * non-zero: value of invalid checksum | |
3273 | * | |
3274 | */ | |
3275 | static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb, | |
3276 | bool complete, | |
3277 | __wsum psum) | |
3278 | { | |
3279 | if (skb->ip_summed == CHECKSUM_COMPLETE) { | |
3280 | if (!csum_fold(csum_add(psum, skb->csum))) { | |
5d0c2b95 | 3281 | skb->csum_valid = 1; |
76ba0aae TH |
3282 | return 0; |
3283 | } | |
5a212329 TH |
3284 | } else if (skb->csum_bad) { |
3285 | /* ip_summed == CHECKSUM_NONE in this case */ | |
c91d4606 | 3286 | return (__force __sum16)1; |
76ba0aae TH |
3287 | } |
3288 | ||
3289 | skb->csum = psum; | |
3290 | ||
5d0c2b95 TH |
3291 | if (complete || skb->len <= CHECKSUM_BREAK) { |
3292 | __sum16 csum; | |
3293 | ||
3294 | csum = __skb_checksum_complete(skb); | |
3295 | skb->csum_valid = !csum; | |
3296 | return csum; | |
3297 | } | |
76ba0aae TH |
3298 | |
3299 | return 0; | |
3300 | } | |
3301 | ||
3302 | static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto) | |
3303 | { | |
3304 | return 0; | |
3305 | } | |
3306 | ||
3307 | /* Perform checksum validate (init). Note that this is a macro since we only | |
3308 | * want to calculate the pseudo header which is an input function if necessary. | |
3309 | * First we try to validate without any computation (checksum unnecessary) and | |
3310 | * then calculate based on checksum complete calling the function to compute | |
3311 | * pseudo header. | |
3312 | * | |
3313 | * Return values: | |
3314 | * 0: checksum is validated or try to in skb_checksum_complete | |
3315 | * non-zero: value of invalid checksum | |
3316 | */ | |
3317 | #define __skb_checksum_validate(skb, proto, complete, \ | |
3318 | zero_okay, check, compute_pseudo) \ | |
3319 | ({ \ | |
3320 | __sum16 __ret = 0; \ | |
5d0c2b95 | 3321 | skb->csum_valid = 0; \ |
76ba0aae TH |
3322 | if (__skb_checksum_validate_needed(skb, zero_okay, check)) \ |
3323 | __ret = __skb_checksum_validate_complete(skb, \ | |
3324 | complete, compute_pseudo(skb, proto)); \ | |
3325 | __ret; \ | |
3326 | }) | |
3327 | ||
3328 | #define skb_checksum_init(skb, proto, compute_pseudo) \ | |
3329 | __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo) | |
3330 | ||
3331 | #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \ | |
3332 | __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo) | |
3333 | ||
3334 | #define skb_checksum_validate(skb, proto, compute_pseudo) \ | |
3335 | __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo) | |
3336 | ||
3337 | #define skb_checksum_validate_zero_check(skb, proto, check, \ | |
3338 | compute_pseudo) \ | |
096a4cfa | 3339 | __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo) |
76ba0aae TH |
3340 | |
3341 | #define skb_checksum_simple_validate(skb) \ | |
3342 | __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo) | |
3343 | ||
d96535a1 TH |
3344 | static inline bool __skb_checksum_convert_check(struct sk_buff *skb) |
3345 | { | |
3346 | return (skb->ip_summed == CHECKSUM_NONE && | |
3347 | skb->csum_valid && !skb->csum_bad); | |
3348 | } | |
3349 | ||
3350 | static inline void __skb_checksum_convert(struct sk_buff *skb, | |
3351 | __sum16 check, __wsum pseudo) | |
3352 | { | |
3353 | skb->csum = ~pseudo; | |
3354 | skb->ip_summed = CHECKSUM_COMPLETE; | |
3355 | } | |
3356 | ||
3357 | #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \ | |
3358 | do { \ | |
3359 | if (__skb_checksum_convert_check(skb)) \ | |
3360 | __skb_checksum_convert(skb, check, \ | |
3361 | compute_pseudo(skb, proto)); \ | |
3362 | } while (0) | |
3363 | ||
15e2396d TH |
3364 | static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr, |
3365 | u16 start, u16 offset) | |
3366 | { | |
3367 | skb->ip_summed = CHECKSUM_PARTIAL; | |
3368 | skb->csum_start = ((unsigned char *)ptr + start) - skb->head; | |
3369 | skb->csum_offset = offset - start; | |
3370 | } | |
3371 | ||
dcdc8994 TH |
3372 | /* Update skbuf and packet to reflect the remote checksum offload operation. |
3373 | * When called, ptr indicates the starting point for skb->csum when | |
3374 | * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete | |
3375 | * here, skb_postpull_rcsum is done so skb->csum start is ptr. | |
3376 | */ | |
3377 | static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr, | |
15e2396d | 3378 | int start, int offset, bool nopartial) |
dcdc8994 TH |
3379 | { |
3380 | __wsum delta; | |
3381 | ||
15e2396d TH |
3382 | if (!nopartial) { |
3383 | skb_remcsum_adjust_partial(skb, ptr, start, offset); | |
3384 | return; | |
3385 | } | |
3386 | ||
dcdc8994 TH |
3387 | if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) { |
3388 | __skb_checksum_complete(skb); | |
3389 | skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data); | |
3390 | } | |
3391 | ||
3392 | delta = remcsum_adjust(ptr, skb->csum, start, offset); | |
3393 | ||
3394 | /* Adjust skb->csum since we changed the packet */ | |
3395 | skb->csum = csum_add(skb->csum, delta); | |
3396 | } | |
3397 | ||
5f79e0f9 | 3398 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) |
7965bd4d | 3399 | void nf_conntrack_destroy(struct nf_conntrack *nfct); |
1da177e4 LT |
3400 | static inline void nf_conntrack_put(struct nf_conntrack *nfct) |
3401 | { | |
3402 | if (nfct && atomic_dec_and_test(&nfct->use)) | |
de6e05c4 | 3403 | nf_conntrack_destroy(nfct); |
1da177e4 LT |
3404 | } |
3405 | static inline void nf_conntrack_get(struct nf_conntrack *nfct) | |
3406 | { | |
3407 | if (nfct) | |
3408 | atomic_inc(&nfct->use); | |
3409 | } | |
2fc72c7b | 3410 | #endif |
34666d46 | 3411 | #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) |
1da177e4 LT |
3412 | static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge) |
3413 | { | |
3414 | if (nf_bridge && atomic_dec_and_test(&nf_bridge->use)) | |
3415 | kfree(nf_bridge); | |
3416 | } | |
3417 | static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge) | |
3418 | { | |
3419 | if (nf_bridge) | |
3420 | atomic_inc(&nf_bridge->use); | |
3421 | } | |
3422 | #endif /* CONFIG_BRIDGE_NETFILTER */ | |
a193a4ab PM |
3423 | static inline void nf_reset(struct sk_buff *skb) |
3424 | { | |
5f79e0f9 | 3425 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) |
a193a4ab PM |
3426 | nf_conntrack_put(skb->nfct); |
3427 | skb->nfct = NULL; | |
2fc72c7b | 3428 | #endif |
34666d46 | 3429 | #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) |
a193a4ab PM |
3430 | nf_bridge_put(skb->nf_bridge); |
3431 | skb->nf_bridge = NULL; | |
3432 | #endif | |
3433 | } | |
3434 | ||
124dff01 PM |
3435 | static inline void nf_reset_trace(struct sk_buff *skb) |
3436 | { | |
478b360a | 3437 | #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES) |
130549fe G |
3438 | skb->nf_trace = 0; |
3439 | #endif | |
a193a4ab PM |
3440 | } |
3441 | ||
edda553c | 3442 | /* Note: This doesn't put any conntrack and bridge info in dst. */ |
b1937227 ED |
3443 | static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src, |
3444 | bool copy) | |
edda553c | 3445 | { |
5f79e0f9 | 3446 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) |
edda553c YK |
3447 | dst->nfct = src->nfct; |
3448 | nf_conntrack_get(src->nfct); | |
b1937227 ED |
3449 | if (copy) |
3450 | dst->nfctinfo = src->nfctinfo; | |
2fc72c7b | 3451 | #endif |
34666d46 | 3452 | #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) |
edda553c YK |
3453 | dst->nf_bridge = src->nf_bridge; |
3454 | nf_bridge_get(src->nf_bridge); | |
3455 | #endif | |
478b360a | 3456 | #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES) |
b1937227 ED |
3457 | if (copy) |
3458 | dst->nf_trace = src->nf_trace; | |
478b360a | 3459 | #endif |
edda553c YK |
3460 | } |
3461 | ||
e7ac05f3 YK |
3462 | static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src) |
3463 | { | |
e7ac05f3 | 3464 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) |
5f79e0f9 | 3465 | nf_conntrack_put(dst->nfct); |
2fc72c7b | 3466 | #endif |
34666d46 | 3467 | #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) |
e7ac05f3 YK |
3468 | nf_bridge_put(dst->nf_bridge); |
3469 | #endif | |
b1937227 | 3470 | __nf_copy(dst, src, true); |
e7ac05f3 YK |
3471 | } |
3472 | ||
984bc16c JM |
3473 | #ifdef CONFIG_NETWORK_SECMARK |
3474 | static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) | |
3475 | { | |
3476 | to->secmark = from->secmark; | |
3477 | } | |
3478 | ||
3479 | static inline void skb_init_secmark(struct sk_buff *skb) | |
3480 | { | |
3481 | skb->secmark = 0; | |
3482 | } | |
3483 | #else | |
3484 | static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) | |
3485 | { } | |
3486 | ||
3487 | static inline void skb_init_secmark(struct sk_buff *skb) | |
3488 | { } | |
3489 | #endif | |
3490 | ||
574f7194 EB |
3491 | static inline bool skb_irq_freeable(const struct sk_buff *skb) |
3492 | { | |
3493 | return !skb->destructor && | |
3494 | #if IS_ENABLED(CONFIG_XFRM) | |
3495 | !skb->sp && | |
3496 | #endif | |
3497 | #if IS_ENABLED(CONFIG_NF_CONNTRACK) | |
3498 | !skb->nfct && | |
3499 | #endif | |
3500 | !skb->_skb_refdst && | |
3501 | !skb_has_frag_list(skb); | |
3502 | } | |
3503 | ||
f25f4e44 PWJ |
3504 | static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping) |
3505 | { | |
f25f4e44 | 3506 | skb->queue_mapping = queue_mapping; |
f25f4e44 PWJ |
3507 | } |
3508 | ||
9247744e | 3509 | static inline u16 skb_get_queue_mapping(const struct sk_buff *skb) |
4e3ab47a | 3510 | { |
4e3ab47a | 3511 | return skb->queue_mapping; |
4e3ab47a PE |
3512 | } |
3513 | ||
f25f4e44 PWJ |
3514 | static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from) |
3515 | { | |
f25f4e44 | 3516 | to->queue_mapping = from->queue_mapping; |
f25f4e44 PWJ |
3517 | } |
3518 | ||
d5a9e24a DM |
3519 | static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue) |
3520 | { | |
3521 | skb->queue_mapping = rx_queue + 1; | |
3522 | } | |
3523 | ||
9247744e | 3524 | static inline u16 skb_get_rx_queue(const struct sk_buff *skb) |
d5a9e24a DM |
3525 | { |
3526 | return skb->queue_mapping - 1; | |
3527 | } | |
3528 | ||
9247744e | 3529 | static inline bool skb_rx_queue_recorded(const struct sk_buff *skb) |
d5a9e24a | 3530 | { |
a02cec21 | 3531 | return skb->queue_mapping != 0; |
d5a9e24a DM |
3532 | } |
3533 | ||
def8b4fa AD |
3534 | static inline struct sec_path *skb_sec_path(struct sk_buff *skb) |
3535 | { | |
0b3d8e08 | 3536 | #ifdef CONFIG_XFRM |
def8b4fa | 3537 | return skb->sp; |
def8b4fa | 3538 | #else |
def8b4fa | 3539 | return NULL; |
def8b4fa | 3540 | #endif |
0b3d8e08 | 3541 | } |
def8b4fa | 3542 | |
68c33163 PS |
3543 | /* Keeps track of mac header offset relative to skb->head. |
3544 | * It is useful for TSO of Tunneling protocol. e.g. GRE. | |
3545 | * For non-tunnel skb it points to skb_mac_header() and for | |
3347c960 ED |
3546 | * tunnel skb it points to outer mac header. |
3547 | * Keeps track of level of encapsulation of network headers. | |
3548 | */ | |
68c33163 | 3549 | struct skb_gso_cb { |
3347c960 ED |
3550 | int mac_offset; |
3551 | int encap_level; | |
7e2b10c1 | 3552 | __u16 csum_start; |
68c33163 | 3553 | }; |
9207f9d4 KK |
3554 | #define SKB_SGO_CB_OFFSET 32 |
3555 | #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET)) | |
68c33163 PS |
3556 | |
3557 | static inline int skb_tnl_header_len(const struct sk_buff *inner_skb) | |
3558 | { | |
3559 | return (skb_mac_header(inner_skb) - inner_skb->head) - | |
3560 | SKB_GSO_CB(inner_skb)->mac_offset; | |
3561 | } | |
3562 | ||
1e2bd517 PS |
3563 | static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra) |
3564 | { | |
3565 | int new_headroom, headroom; | |
3566 | int ret; | |
3567 | ||
3568 | headroom = skb_headroom(skb); | |
3569 | ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC); | |
3570 | if (ret) | |
3571 | return ret; | |
3572 | ||
3573 | new_headroom = skb_headroom(skb); | |
3574 | SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom); | |
3575 | return 0; | |
3576 | } | |
3577 | ||
7e2b10c1 TH |
3578 | /* Compute the checksum for a gso segment. First compute the checksum value |
3579 | * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and | |
3580 | * then add in skb->csum (checksum from csum_start to end of packet). | |
3581 | * skb->csum and csum_start are then updated to reflect the checksum of the | |
3582 | * resultant packet starting from the transport header-- the resultant checksum | |
3583 | * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo | |
3584 | * header. | |
3585 | */ | |
3586 | static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res) | |
3587 | { | |
3588 | int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) - | |
c91d4606 ED |
3589 | skb_transport_offset(skb); |
3590 | __wsum partial; | |
7e2b10c1 | 3591 | |
c91d4606 | 3592 | partial = csum_partial(skb_transport_header(skb), plen, skb->csum); |
7e2b10c1 TH |
3593 | skb->csum = res; |
3594 | SKB_GSO_CB(skb)->csum_start -= plen; | |
3595 | ||
c91d4606 | 3596 | return csum_fold(partial); |
7e2b10c1 TH |
3597 | } |
3598 | ||
bdcc0924 | 3599 | static inline bool skb_is_gso(const struct sk_buff *skb) |
89114afd HX |
3600 | { |
3601 | return skb_shinfo(skb)->gso_size; | |
3602 | } | |
3603 | ||
36a8f39e | 3604 | /* Note: Should be called only if skb_is_gso(skb) is true */ |
bdcc0924 | 3605 | static inline bool skb_is_gso_v6(const struct sk_buff *skb) |
eabd7e35 BG |
3606 | { |
3607 | return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6; | |
3608 | } | |
3609 | ||
7965bd4d | 3610 | void __skb_warn_lro_forwarding(const struct sk_buff *skb); |
4497b076 BH |
3611 | |
3612 | static inline bool skb_warn_if_lro(const struct sk_buff *skb) | |
3613 | { | |
3614 | /* LRO sets gso_size but not gso_type, whereas if GSO is really | |
3615 | * wanted then gso_type will be set. */ | |
05bdd2f1 ED |
3616 | const struct skb_shared_info *shinfo = skb_shinfo(skb); |
3617 | ||
b78462eb AD |
3618 | if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 && |
3619 | unlikely(shinfo->gso_type == 0)) { | |
4497b076 BH |
3620 | __skb_warn_lro_forwarding(skb); |
3621 | return true; | |
3622 | } | |
3623 | return false; | |
3624 | } | |
3625 | ||
35fc92a9 HX |
3626 | static inline void skb_forward_csum(struct sk_buff *skb) |
3627 | { | |
3628 | /* Unfortunately we don't support this one. Any brave souls? */ | |
3629 | if (skb->ip_summed == CHECKSUM_COMPLETE) | |
3630 | skb->ip_summed = CHECKSUM_NONE; | |
3631 | } | |
3632 | ||
bc8acf2c ED |
3633 | /** |
3634 | * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE | |
3635 | * @skb: skb to check | |
3636 | * | |
3637 | * fresh skbs have their ip_summed set to CHECKSUM_NONE. | |
3638 | * Instead of forcing ip_summed to CHECKSUM_NONE, we can | |
3639 | * use this helper, to document places where we make this assertion. | |
3640 | */ | |
05bdd2f1 | 3641 | static inline void skb_checksum_none_assert(const struct sk_buff *skb) |
bc8acf2c ED |
3642 | { |
3643 | #ifdef DEBUG | |
3644 | BUG_ON(skb->ip_summed != CHECKSUM_NONE); | |
3645 | #endif | |
3646 | } | |
3647 | ||
f35d9d8a | 3648 | bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off); |
a6686f2f | 3649 | |
ed1f50c3 | 3650 | int skb_checksum_setup(struct sk_buff *skb, bool recalculate); |
9afd85c9 LL |
3651 | struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb, |
3652 | unsigned int transport_len, | |
3653 | __sum16(*skb_chkf)(struct sk_buff *skb)); | |
ed1f50c3 | 3654 | |
3a7c1ee4 AD |
3655 | /** |
3656 | * skb_head_is_locked - Determine if the skb->head is locked down | |
3657 | * @skb: skb to check | |
3658 | * | |
3659 | * The head on skbs build around a head frag can be removed if they are | |
3660 | * not cloned. This function returns true if the skb head is locked down | |
3661 | * due to either being allocated via kmalloc, or by being a clone with | |
3662 | * multiple references to the head. | |
3663 | */ | |
3664 | static inline bool skb_head_is_locked(const struct sk_buff *skb) | |
3665 | { | |
3666 | return !skb->head_frag || skb_cloned(skb); | |
3667 | } | |
fe6cc55f FW |
3668 | |
3669 | /** | |
3670 | * skb_gso_network_seglen - Return length of individual segments of a gso packet | |
3671 | * | |
3672 | * @skb: GSO skb | |
3673 | * | |
3674 | * skb_gso_network_seglen is used to determine the real size of the | |
3675 | * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP). | |
3676 | * | |
3677 | * The MAC/L2 header is not accounted for. | |
3678 | */ | |
3679 | static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb) | |
3680 | { | |
3681 | unsigned int hdr_len = skb_transport_header(skb) - | |
3682 | skb_network_header(skb); | |
3683 | return hdr_len + skb_gso_transport_seglen(skb); | |
3684 | } | |
ee122c79 | 3685 | |
1da177e4 LT |
3686 | #endif /* __KERNEL__ */ |
3687 | #endif /* _LINUX_SKBUFF_H */ |