]>
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 |
5f74f82e | 302 | extern int sysctl_max_skb_frags; |
1da177e4 | 303 | |
3953c46c MRL |
304 | /* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to |
305 | * segment using its current segmentation instead. | |
306 | */ | |
307 | #define GSO_BY_FRAGS 0xFFFF | |
308 | ||
1da177e4 LT |
309 | typedef struct skb_frag_struct skb_frag_t; |
310 | ||
311 | struct skb_frag_struct { | |
a8605c60 IC |
312 | struct { |
313 | struct page *p; | |
314 | } page; | |
cb4dfe56 | 315 | #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536) |
a309bb07 DM |
316 | __u32 page_offset; |
317 | __u32 size; | |
cb4dfe56 ED |
318 | #else |
319 | __u16 page_offset; | |
320 | __u16 size; | |
321 | #endif | |
1da177e4 LT |
322 | }; |
323 | ||
9e903e08 ED |
324 | static inline unsigned int skb_frag_size(const skb_frag_t *frag) |
325 | { | |
326 | return frag->size; | |
327 | } | |
328 | ||
329 | static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size) | |
330 | { | |
331 | frag->size = size; | |
332 | } | |
333 | ||
334 | static inline void skb_frag_size_add(skb_frag_t *frag, int delta) | |
335 | { | |
336 | frag->size += delta; | |
337 | } | |
338 | ||
339 | static inline void skb_frag_size_sub(skb_frag_t *frag, int delta) | |
340 | { | |
341 | frag->size -= delta; | |
342 | } | |
343 | ||
ac45f602 PO |
344 | #define HAVE_HW_TIME_STAMP |
345 | ||
346 | /** | |
d3a21be8 | 347 | * struct skb_shared_hwtstamps - hardware time stamps |
ac45f602 PO |
348 | * @hwtstamp: hardware time stamp transformed into duration |
349 | * since arbitrary point in time | |
ac45f602 PO |
350 | * |
351 | * Software time stamps generated by ktime_get_real() are stored in | |
4d276eb6 | 352 | * skb->tstamp. |
ac45f602 PO |
353 | * |
354 | * hwtstamps can only be compared against other hwtstamps from | |
355 | * the same device. | |
356 | * | |
357 | * This structure is attached to packets as part of the | |
358 | * &skb_shared_info. Use skb_hwtstamps() to get a pointer. | |
359 | */ | |
360 | struct skb_shared_hwtstamps { | |
361 | ktime_t hwtstamp; | |
ac45f602 PO |
362 | }; |
363 | ||
2244d07b OH |
364 | /* Definitions for tx_flags in struct skb_shared_info */ |
365 | enum { | |
366 | /* generate hardware time stamp */ | |
367 | SKBTX_HW_TSTAMP = 1 << 0, | |
368 | ||
e7fd2885 | 369 | /* generate software time stamp when queueing packet to NIC */ |
2244d07b OH |
370 | SKBTX_SW_TSTAMP = 1 << 1, |
371 | ||
372 | /* device driver is going to provide hardware time stamp */ | |
373 | SKBTX_IN_PROGRESS = 1 << 2, | |
374 | ||
a6686f2f | 375 | /* device driver supports TX zero-copy buffers */ |
62b1a8ab | 376 | SKBTX_DEV_ZEROCOPY = 1 << 3, |
6e3e939f JB |
377 | |
378 | /* generate wifi status information (where possible) */ | |
62b1a8ab | 379 | SKBTX_WIFI_STATUS = 1 << 4, |
c9af6db4 PS |
380 | |
381 | /* This indicates at least one fragment might be overwritten | |
382 | * (as in vmsplice(), sendfile() ...) | |
383 | * If we need to compute a TX checksum, we'll need to copy | |
384 | * all frags to avoid possible bad checksum | |
385 | */ | |
386 | SKBTX_SHARED_FRAG = 1 << 5, | |
e7fd2885 WB |
387 | |
388 | /* generate software time stamp when entering packet scheduling */ | |
389 | SKBTX_SCHED_TSTAMP = 1 << 6, | |
a6686f2f SM |
390 | }; |
391 | ||
e1c8a607 | 392 | #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \ |
0a2cf20c | 393 | SKBTX_SCHED_TSTAMP) |
f24b9be5 WB |
394 | #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP) |
395 | ||
a6686f2f SM |
396 | /* |
397 | * The callback notifies userspace to release buffers when skb DMA is done in | |
398 | * lower device, the skb last reference should be 0 when calling this. | |
e19d6763 MT |
399 | * The zerocopy_success argument is true if zero copy transmit occurred, |
400 | * false on data copy or out of memory error caused by data copy attempt. | |
ca8f4fb2 MT |
401 | * The ctx field is used to track device context. |
402 | * The desc field is used to track userspace buffer index. | |
a6686f2f SM |
403 | */ |
404 | struct ubuf_info { | |
e19d6763 | 405 | void (*callback)(struct ubuf_info *, bool zerocopy_success); |
ca8f4fb2 | 406 | void *ctx; |
a6686f2f | 407 | unsigned long desc; |
ac45f602 PO |
408 | }; |
409 | ||
1da177e4 LT |
410 | /* This data is invariant across clones and lives at |
411 | * the end of the header data, ie. at skb->end. | |
412 | */ | |
413 | struct skb_shared_info { | |
9f42f126 IC |
414 | unsigned char nr_frags; |
415 | __u8 tx_flags; | |
7967168c HX |
416 | unsigned short gso_size; |
417 | /* Warning: this field is not always filled in (UFO)! */ | |
418 | unsigned short gso_segs; | |
419 | unsigned short gso_type; | |
1da177e4 | 420 | struct sk_buff *frag_list; |
ac45f602 | 421 | struct skb_shared_hwtstamps hwtstamps; |
09c2d251 | 422 | u32 tskey; |
9f42f126 | 423 | __be32 ip6_frag_id; |
ec7d2f2c ED |
424 | |
425 | /* | |
426 | * Warning : all fields before dataref are cleared in __alloc_skb() | |
427 | */ | |
428 | atomic_t dataref; | |
429 | ||
69e3c75f JB |
430 | /* Intermediate layers must ensure that destructor_arg |
431 | * remains valid until skb destructor */ | |
432 | void * destructor_arg; | |
a6686f2f | 433 | |
fed66381 ED |
434 | /* must be last field, see pskb_expand_head() */ |
435 | skb_frag_t frags[MAX_SKB_FRAGS]; | |
1da177e4 LT |
436 | }; |
437 | ||
438 | /* We divide dataref into two halves. The higher 16 bits hold references | |
439 | * to the payload part of skb->data. The lower 16 bits hold references to | |
334a8132 PM |
440 | * the entire skb->data. A clone of a headerless skb holds the length of |
441 | * the header in skb->hdr_len. | |
1da177e4 LT |
442 | * |
443 | * All users must obey the rule that the skb->data reference count must be | |
444 | * greater than or equal to the payload reference count. | |
445 | * | |
446 | * Holding a reference to the payload part means that the user does not | |
447 | * care about modifications to the header part of skb->data. | |
448 | */ | |
449 | #define SKB_DATAREF_SHIFT 16 | |
450 | #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1) | |
451 | ||
d179cd12 DM |
452 | |
453 | enum { | |
c8753d55 VS |
454 | SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */ |
455 | SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */ | |
456 | SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */ | |
d179cd12 DM |
457 | }; |
458 | ||
7967168c HX |
459 | enum { |
460 | SKB_GSO_TCPV4 = 1 << 0, | |
f83ef8c0 | 461 | SKB_GSO_UDP = 1 << 1, |
576a30eb HX |
462 | |
463 | /* This indicates the skb is from an untrusted source. */ | |
464 | SKB_GSO_DODGY = 1 << 2, | |
b0da8537 MC |
465 | |
466 | /* This indicates the tcp segment has CWR set. */ | |
f83ef8c0 HX |
467 | SKB_GSO_TCP_ECN = 1 << 3, |
468 | ||
cbc53e08 | 469 | SKB_GSO_TCP_FIXEDID = 1 << 4, |
01d5b2fc | 470 | |
cbc53e08 | 471 | SKB_GSO_TCPV6 = 1 << 5, |
68c33163 | 472 | |
cbc53e08 | 473 | SKB_GSO_FCOE = 1 << 6, |
73136267 | 474 | |
cbc53e08 | 475 | SKB_GSO_GRE = 1 << 7, |
0d89d203 | 476 | |
cbc53e08 | 477 | SKB_GSO_GRE_CSUM = 1 << 8, |
cb32f511 | 478 | |
7e13318d | 479 | SKB_GSO_IPXIP4 = 1 << 9, |
61c1db7f | 480 | |
7e13318d | 481 | SKB_GSO_IPXIP6 = 1 << 10, |
0f4f4ffa | 482 | |
cbc53e08 | 483 | SKB_GSO_UDP_TUNNEL = 1 << 11, |
4749c09c | 484 | |
cbc53e08 AD |
485 | SKB_GSO_UDP_TUNNEL_CSUM = 1 << 12, |
486 | ||
802ab55a AD |
487 | SKB_GSO_PARTIAL = 1 << 13, |
488 | ||
489 | SKB_GSO_TUNNEL_REMCSUM = 1 << 14, | |
7967168c HX |
490 | }; |
491 | ||
2e07fa9c ACM |
492 | #if BITS_PER_LONG > 32 |
493 | #define NET_SKBUFF_DATA_USES_OFFSET 1 | |
494 | #endif | |
495 | ||
496 | #ifdef NET_SKBUFF_DATA_USES_OFFSET | |
497 | typedef unsigned int sk_buff_data_t; | |
498 | #else | |
499 | typedef unsigned char *sk_buff_data_t; | |
500 | #endif | |
501 | ||
363ec392 ED |
502 | /** |
503 | * struct skb_mstamp - multi resolution time stamps | |
504 | * @stamp_us: timestamp in us resolution | |
505 | * @stamp_jiffies: timestamp in jiffies | |
506 | */ | |
507 | struct skb_mstamp { | |
508 | union { | |
509 | u64 v64; | |
510 | struct { | |
511 | u32 stamp_us; | |
512 | u32 stamp_jiffies; | |
513 | }; | |
514 | }; | |
515 | }; | |
516 | ||
517 | /** | |
518 | * skb_mstamp_get - get current timestamp | |
519 | * @cl: place to store timestamps | |
520 | */ | |
521 | static inline void skb_mstamp_get(struct skb_mstamp *cl) | |
522 | { | |
523 | u64 val = local_clock(); | |
524 | ||
525 | do_div(val, NSEC_PER_USEC); | |
526 | cl->stamp_us = (u32)val; | |
527 | cl->stamp_jiffies = (u32)jiffies; | |
528 | } | |
529 | ||
530 | /** | |
531 | * skb_mstamp_delta - compute the difference in usec between two skb_mstamp | |
532 | * @t1: pointer to newest sample | |
533 | * @t0: pointer to oldest sample | |
534 | */ | |
535 | static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1, | |
536 | const struct skb_mstamp *t0) | |
537 | { | |
538 | s32 delta_us = t1->stamp_us - t0->stamp_us; | |
539 | u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies; | |
540 | ||
541 | /* If delta_us is negative, this might be because interval is too big, | |
542 | * or local_clock() drift is too big : fallback using jiffies. | |
543 | */ | |
544 | if (delta_us <= 0 || | |
545 | delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ))) | |
546 | ||
547 | delta_us = jiffies_to_usecs(delta_jiffies); | |
548 | ||
549 | return delta_us; | |
550 | } | |
551 | ||
625a5e10 YC |
552 | static inline bool skb_mstamp_after(const struct skb_mstamp *t1, |
553 | const struct skb_mstamp *t0) | |
554 | { | |
555 | s32 diff = t1->stamp_jiffies - t0->stamp_jiffies; | |
556 | ||
557 | if (!diff) | |
558 | diff = t1->stamp_us - t0->stamp_us; | |
559 | return diff > 0; | |
560 | } | |
363ec392 | 561 | |
1da177e4 LT |
562 | /** |
563 | * struct sk_buff - socket buffer | |
564 | * @next: Next buffer in list | |
565 | * @prev: Previous buffer in list | |
363ec392 | 566 | * @tstamp: Time we arrived/left |
56b17425 | 567 | * @rbnode: RB tree node, alternative to next/prev for netem/tcp |
d84e0bd7 | 568 | * @sk: Socket we are owned by |
1da177e4 | 569 | * @dev: Device we arrived on/are leaving by |
d84e0bd7 | 570 | * @cb: Control buffer. Free for use by every layer. Put private vars here |
7fee226a | 571 | * @_skb_refdst: destination entry (with norefcount bit) |
67be2dd1 | 572 | * @sp: the security path, used for xfrm |
1da177e4 LT |
573 | * @len: Length of actual data |
574 | * @data_len: Data length | |
575 | * @mac_len: Length of link layer header | |
334a8132 | 576 | * @hdr_len: writable header length of cloned skb |
663ead3b HX |
577 | * @csum: Checksum (must include start/offset pair) |
578 | * @csum_start: Offset from skb->head where checksumming should start | |
579 | * @csum_offset: Offset from csum_start where checksum should be stored | |
d84e0bd7 | 580 | * @priority: Packet queueing priority |
60ff7467 | 581 | * @ignore_df: allow local fragmentation |
1da177e4 | 582 | * @cloned: Head may be cloned (check refcnt to be sure) |
d84e0bd7 | 583 | * @ip_summed: Driver fed us an IP checksum |
1da177e4 | 584 | * @nohdr: Payload reference only, must not modify header |
d84e0bd7 | 585 | * @nfctinfo: Relationship of this skb to the connection |
1da177e4 | 586 | * @pkt_type: Packet class |
c83c2486 | 587 | * @fclone: skbuff clone status |
c83c2486 | 588 | * @ipvs_property: skbuff is owned by ipvs |
31729363 RD |
589 | * @peeked: this packet has been seen already, so stats have been |
590 | * done for it, don't do them again | |
ba9dda3a | 591 | * @nf_trace: netfilter packet trace flag |
d84e0bd7 DB |
592 | * @protocol: Packet protocol from driver |
593 | * @destructor: Destruct function | |
594 | * @nfct: Associated connection, if any | |
1da177e4 | 595 | * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c |
8964be4a | 596 | * @skb_iif: ifindex of device we arrived on |
1da177e4 LT |
597 | * @tc_index: Traffic control index |
598 | * @tc_verd: traffic control verdict | |
61b905da | 599 | * @hash: the packet hash |
d84e0bd7 | 600 | * @queue_mapping: Queue mapping for multiqueue devices |
0b725a2c | 601 | * @xmit_more: More SKBs are pending for this queue |
553a5672 | 602 | * @ndisc_nodetype: router type (from link layer) |
d84e0bd7 | 603 | * @ooo_okay: allow the mapping of a socket to a queue to be changed |
61b905da | 604 | * @l4_hash: indicate hash is a canonical 4-tuple hash over transport |
4ca2462e | 605 | * ports. |
a3b18ddb | 606 | * @sw_hash: indicates hash was computed in software stack |
6e3e939f JB |
607 | * @wifi_acked_valid: wifi_acked was set |
608 | * @wifi_acked: whether frame was acked on wifi or not | |
3bdc0eba | 609 | * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS |
06021292 | 610 | * @napi_id: id of the NAPI struct this skb came from |
984bc16c | 611 | * @secmark: security marking |
0c4f691f | 612 | * @offload_fwd_mark: fwding offload mark |
d84e0bd7 | 613 | * @mark: Generic packet mark |
86a9bad3 | 614 | * @vlan_proto: vlan encapsulation protocol |
6aa895b0 | 615 | * @vlan_tci: vlan tag control information |
0d89d203 | 616 | * @inner_protocol: Protocol (encapsulation) |
6a674e9c JG |
617 | * @inner_transport_header: Inner transport layer header (encapsulation) |
618 | * @inner_network_header: Network layer header (encapsulation) | |
aefbd2b3 | 619 | * @inner_mac_header: Link layer header (encapsulation) |
d84e0bd7 DB |
620 | * @transport_header: Transport layer header |
621 | * @network_header: Network layer header | |
622 | * @mac_header: Link layer header | |
623 | * @tail: Tail pointer | |
624 | * @end: End pointer | |
625 | * @head: Head of buffer | |
626 | * @data: Data head pointer | |
627 | * @truesize: Buffer size | |
628 | * @users: User count - see {datagram,tcp}.c | |
1da177e4 LT |
629 | */ |
630 | ||
631 | struct sk_buff { | |
363ec392 | 632 | union { |
56b17425 ED |
633 | struct { |
634 | /* These two members must be first. */ | |
635 | struct sk_buff *next; | |
636 | struct sk_buff *prev; | |
637 | ||
638 | union { | |
639 | ktime_t tstamp; | |
640 | struct skb_mstamp skb_mstamp; | |
641 | }; | |
642 | }; | |
643 | struct rb_node rbnode; /* used in netem & tcp stack */ | |
363ec392 | 644 | }; |
da3f5cf1 | 645 | struct sock *sk; |
1da177e4 | 646 | struct net_device *dev; |
1da177e4 | 647 | |
1da177e4 LT |
648 | /* |
649 | * This is the control buffer. It is free to use for every | |
650 | * layer. Please put your private variables there. If you | |
651 | * want to keep them across layers you have to do a skb_clone() | |
652 | * first. This is owned by whoever has the skb queued ATM. | |
653 | */ | |
da3f5cf1 | 654 | char cb[48] __aligned(8); |
1da177e4 | 655 | |
7fee226a | 656 | unsigned long _skb_refdst; |
b1937227 | 657 | void (*destructor)(struct sk_buff *skb); |
da3f5cf1 FF |
658 | #ifdef CONFIG_XFRM |
659 | struct sec_path *sp; | |
b1937227 ED |
660 | #endif |
661 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) | |
662 | struct nf_conntrack *nfct; | |
663 | #endif | |
85224844 | 664 | #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) |
b1937227 | 665 | struct nf_bridge_info *nf_bridge; |
da3f5cf1 | 666 | #endif |
1da177e4 | 667 | unsigned int len, |
334a8132 PM |
668 | data_len; |
669 | __u16 mac_len, | |
670 | hdr_len; | |
b1937227 ED |
671 | |
672 | /* Following fields are _not_ copied in __copy_skb_header() | |
673 | * Note that queue_mapping is here mostly to fill a hole. | |
674 | */ | |
fe55f6d5 | 675 | kmemcheck_bitfield_begin(flags1); |
b1937227 ED |
676 | __u16 queue_mapping; |
677 | __u8 cloned:1, | |
6869c4d8 | 678 | nohdr:1, |
b84f4cc9 | 679 | fclone:2, |
a59322be | 680 | peeked:1, |
b1937227 ED |
681 | head_frag:1, |
682 | xmit_more:1; | |
683 | /* one bit hole */ | |
fe55f6d5 | 684 | kmemcheck_bitfield_end(flags1); |
4031ae6e | 685 | |
b1937227 ED |
686 | /* fields enclosed in headers_start/headers_end are copied |
687 | * using a single memcpy() in __copy_skb_header() | |
688 | */ | |
ebcf34f3 | 689 | /* private: */ |
b1937227 | 690 | __u32 headers_start[0]; |
ebcf34f3 | 691 | /* public: */ |
4031ae6e | 692 | |
233577a2 HFS |
693 | /* if you move pkt_type around you also must adapt those constants */ |
694 | #ifdef __BIG_ENDIAN_BITFIELD | |
695 | #define PKT_TYPE_MAX (7 << 5) | |
696 | #else | |
697 | #define PKT_TYPE_MAX 7 | |
1da177e4 | 698 | #endif |
233577a2 | 699 | #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset) |
fe55f6d5 | 700 | |
233577a2 | 701 | __u8 __pkt_type_offset[0]; |
b1937227 | 702 | __u8 pkt_type:3; |
c93bdd0e | 703 | __u8 pfmemalloc:1; |
b1937227 ED |
704 | __u8 ignore_df:1; |
705 | __u8 nfctinfo:3; | |
706 | ||
707 | __u8 nf_trace:1; | |
708 | __u8 ip_summed:2; | |
3853b584 | 709 | __u8 ooo_okay:1; |
61b905da | 710 | __u8 l4_hash:1; |
a3b18ddb | 711 | __u8 sw_hash:1; |
6e3e939f JB |
712 | __u8 wifi_acked_valid:1; |
713 | __u8 wifi_acked:1; | |
b1937227 | 714 | |
3bdc0eba | 715 | __u8 no_fcs:1; |
77cffe23 | 716 | /* Indicates the inner headers are valid in the skbuff. */ |
6a674e9c | 717 | __u8 encapsulation:1; |
7e2b10c1 | 718 | __u8 encap_hdr_csum:1; |
5d0c2b95 | 719 | __u8 csum_valid:1; |
7e3cead5 | 720 | __u8 csum_complete_sw:1; |
b1937227 ED |
721 | __u8 csum_level:2; |
722 | __u8 csum_bad:1; | |
fe55f6d5 | 723 | |
b1937227 ED |
724 | #ifdef CONFIG_IPV6_NDISC_NODETYPE |
725 | __u8 ndisc_nodetype:2; | |
726 | #endif | |
727 | __u8 ipvs_property:1; | |
8bce6d7d | 728 | __u8 inner_protocol_type:1; |
e585f236 TH |
729 | __u8 remcsum_offload:1; |
730 | /* 3 or 5 bit hole */ | |
b1937227 ED |
731 | |
732 | #ifdef CONFIG_NET_SCHED | |
733 | __u16 tc_index; /* traffic control index */ | |
734 | #ifdef CONFIG_NET_CLS_ACT | |
735 | __u16 tc_verd; /* traffic control verdict */ | |
736 | #endif | |
737 | #endif | |
fe55f6d5 | 738 | |
b1937227 ED |
739 | union { |
740 | __wsum csum; | |
741 | struct { | |
742 | __u16 csum_start; | |
743 | __u16 csum_offset; | |
744 | }; | |
745 | }; | |
746 | __u32 priority; | |
747 | int skb_iif; | |
748 | __u32 hash; | |
749 | __be16 vlan_proto; | |
750 | __u16 vlan_tci; | |
2bd82484 ED |
751 | #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS) |
752 | union { | |
753 | unsigned int napi_id; | |
754 | unsigned int sender_cpu; | |
755 | }; | |
97fc2f08 | 756 | #endif |
0c4f691f | 757 | union { |
984bc16c | 758 | #ifdef CONFIG_NETWORK_SECMARK |
0c4f691f SF |
759 | __u32 secmark; |
760 | #endif | |
761 | #ifdef CONFIG_NET_SWITCHDEV | |
762 | __u32 offload_fwd_mark; | |
984bc16c | 763 | #endif |
0c4f691f SF |
764 | }; |
765 | ||
3b885787 NH |
766 | union { |
767 | __u32 mark; | |
16fad69c | 768 | __u32 reserved_tailroom; |
3b885787 | 769 | }; |
1da177e4 | 770 | |
8bce6d7d TH |
771 | union { |
772 | __be16 inner_protocol; | |
773 | __u8 inner_ipproto; | |
774 | }; | |
775 | ||
1a37e412 SH |
776 | __u16 inner_transport_header; |
777 | __u16 inner_network_header; | |
778 | __u16 inner_mac_header; | |
b1937227 ED |
779 | |
780 | __be16 protocol; | |
1a37e412 SH |
781 | __u16 transport_header; |
782 | __u16 network_header; | |
783 | __u16 mac_header; | |
b1937227 | 784 | |
ebcf34f3 | 785 | /* private: */ |
b1937227 | 786 | __u32 headers_end[0]; |
ebcf34f3 | 787 | /* public: */ |
b1937227 | 788 | |
1da177e4 | 789 | /* These elements must be at the end, see alloc_skb() for details. */ |
27a884dc | 790 | sk_buff_data_t tail; |
4305b541 | 791 | sk_buff_data_t end; |
1da177e4 | 792 | unsigned char *head, |
4305b541 | 793 | *data; |
27a884dc ACM |
794 | unsigned int truesize; |
795 | atomic_t users; | |
1da177e4 LT |
796 | }; |
797 | ||
798 | #ifdef __KERNEL__ | |
799 | /* | |
800 | * Handling routines are only of interest to the kernel | |
801 | */ | |
802 | #include <linux/slab.h> | |
803 | ||
1da177e4 | 804 | |
c93bdd0e MG |
805 | #define SKB_ALLOC_FCLONE 0x01 |
806 | #define SKB_ALLOC_RX 0x02 | |
fd11a83d | 807 | #define SKB_ALLOC_NAPI 0x04 |
c93bdd0e MG |
808 | |
809 | /* Returns true if the skb was allocated from PFMEMALLOC reserves */ | |
810 | static inline bool skb_pfmemalloc(const struct sk_buff *skb) | |
811 | { | |
812 | return unlikely(skb->pfmemalloc); | |
813 | } | |
814 | ||
7fee226a ED |
815 | /* |
816 | * skb might have a dst pointer attached, refcounted or not. | |
817 | * _skb_refdst low order bit is set if refcount was _not_ taken | |
818 | */ | |
819 | #define SKB_DST_NOREF 1UL | |
820 | #define SKB_DST_PTRMASK ~(SKB_DST_NOREF) | |
821 | ||
822 | /** | |
823 | * skb_dst - returns skb dst_entry | |
824 | * @skb: buffer | |
825 | * | |
826 | * Returns skb dst_entry, regardless of reference taken or not. | |
827 | */ | |
adf30907 ED |
828 | static inline struct dst_entry *skb_dst(const struct sk_buff *skb) |
829 | { | |
7fee226a ED |
830 | /* If refdst was not refcounted, check we still are in a |
831 | * rcu_read_lock section | |
832 | */ | |
833 | WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) && | |
834 | !rcu_read_lock_held() && | |
835 | !rcu_read_lock_bh_held()); | |
836 | return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK); | |
adf30907 ED |
837 | } |
838 | ||
7fee226a ED |
839 | /** |
840 | * skb_dst_set - sets skb dst | |
841 | * @skb: buffer | |
842 | * @dst: dst entry | |
843 | * | |
844 | * Sets skb dst, assuming a reference was taken on dst and should | |
845 | * be released by skb_dst_drop() | |
846 | */ | |
adf30907 ED |
847 | static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst) |
848 | { | |
7fee226a ED |
849 | skb->_skb_refdst = (unsigned long)dst; |
850 | } | |
851 | ||
932bc4d7 JA |
852 | /** |
853 | * skb_dst_set_noref - sets skb dst, hopefully, without taking reference | |
854 | * @skb: buffer | |
855 | * @dst: dst entry | |
856 | * | |
857 | * Sets skb dst, assuming a reference was not taken on dst. | |
858 | * If dst entry is cached, we do not take reference and dst_release | |
859 | * will be avoided by refdst_drop. If dst entry is not cached, we take | |
860 | * reference, so that last dst_release can destroy the dst immediately. | |
861 | */ | |
862 | static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst) | |
863 | { | |
dbfc4fb7 HFS |
864 | WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held()); |
865 | skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF; | |
932bc4d7 | 866 | } |
7fee226a ED |
867 | |
868 | /** | |
25985edc | 869 | * skb_dst_is_noref - Test if skb dst isn't refcounted |
7fee226a ED |
870 | * @skb: buffer |
871 | */ | |
872 | static inline bool skb_dst_is_noref(const struct sk_buff *skb) | |
873 | { | |
874 | return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb); | |
adf30907 ED |
875 | } |
876 | ||
511c3f92 ED |
877 | static inline struct rtable *skb_rtable(const struct sk_buff *skb) |
878 | { | |
adf30907 | 879 | return (struct rtable *)skb_dst(skb); |
511c3f92 ED |
880 | } |
881 | ||
7965bd4d JP |
882 | void kfree_skb(struct sk_buff *skb); |
883 | void kfree_skb_list(struct sk_buff *segs); | |
884 | void skb_tx_error(struct sk_buff *skb); | |
885 | void consume_skb(struct sk_buff *skb); | |
886 | void __kfree_skb(struct sk_buff *skb); | |
d7e8883c | 887 | extern struct kmem_cache *skbuff_head_cache; |
bad43ca8 | 888 | |
7965bd4d JP |
889 | void kfree_skb_partial(struct sk_buff *skb, bool head_stolen); |
890 | bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from, | |
891 | bool *fragstolen, int *delta_truesize); | |
bad43ca8 | 892 | |
7965bd4d JP |
893 | struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags, |
894 | int node); | |
2ea2f62c | 895 | struct sk_buff *__build_skb(void *data, unsigned int frag_size); |
7965bd4d | 896 | struct sk_buff *build_skb(void *data, unsigned int frag_size); |
d179cd12 | 897 | static inline struct sk_buff *alloc_skb(unsigned int size, |
dd0fc66f | 898 | gfp_t priority) |
d179cd12 | 899 | { |
564824b0 | 900 | return __alloc_skb(size, priority, 0, NUMA_NO_NODE); |
d179cd12 DM |
901 | } |
902 | ||
2e4e4410 ED |
903 | struct sk_buff *alloc_skb_with_frags(unsigned long header_len, |
904 | unsigned long data_len, | |
905 | int max_page_order, | |
906 | int *errcode, | |
907 | gfp_t gfp_mask); | |
908 | ||
d0bf4a9e ED |
909 | /* Layout of fast clones : [skb1][skb2][fclone_ref] */ |
910 | struct sk_buff_fclones { | |
911 | struct sk_buff skb1; | |
912 | ||
913 | struct sk_buff skb2; | |
914 | ||
915 | atomic_t fclone_ref; | |
916 | }; | |
917 | ||
918 | /** | |
919 | * skb_fclone_busy - check if fclone is busy | |
920 | * @skb: buffer | |
921 | * | |
bda13fed | 922 | * Returns true if skb is a fast clone, and its clone is not freed. |
39bb5e62 ED |
923 | * Some drivers call skb_orphan() in their ndo_start_xmit(), |
924 | * so we also check that this didnt happen. | |
d0bf4a9e | 925 | */ |
39bb5e62 ED |
926 | static inline bool skb_fclone_busy(const struct sock *sk, |
927 | const struct sk_buff *skb) | |
d0bf4a9e ED |
928 | { |
929 | const struct sk_buff_fclones *fclones; | |
930 | ||
931 | fclones = container_of(skb, struct sk_buff_fclones, skb1); | |
932 | ||
933 | return skb->fclone == SKB_FCLONE_ORIG && | |
6ffe75eb | 934 | atomic_read(&fclones->fclone_ref) > 1 && |
39bb5e62 | 935 | fclones->skb2.sk == sk; |
d0bf4a9e ED |
936 | } |
937 | ||
d179cd12 | 938 | static inline struct sk_buff *alloc_skb_fclone(unsigned int size, |
dd0fc66f | 939 | gfp_t priority) |
d179cd12 | 940 | { |
c93bdd0e | 941 | return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE); |
d179cd12 DM |
942 | } |
943 | ||
7965bd4d | 944 | struct sk_buff *__alloc_skb_head(gfp_t priority, int node); |
0ebd0ac5 PM |
945 | static inline struct sk_buff *alloc_skb_head(gfp_t priority) |
946 | { | |
947 | return __alloc_skb_head(priority, -1); | |
948 | } | |
949 | ||
7965bd4d JP |
950 | struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src); |
951 | int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask); | |
952 | struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority); | |
953 | struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority); | |
bad93e9d OP |
954 | struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom, |
955 | gfp_t gfp_mask, bool fclone); | |
956 | static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom, | |
957 | gfp_t gfp_mask) | |
958 | { | |
959 | return __pskb_copy_fclone(skb, headroom, gfp_mask, false); | |
960 | } | |
7965bd4d JP |
961 | |
962 | int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask); | |
963 | struct sk_buff *skb_realloc_headroom(struct sk_buff *skb, | |
964 | unsigned int headroom); | |
965 | struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom, | |
966 | int newtailroom, gfp_t priority); | |
25a91d8d FD |
967 | int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg, |
968 | int offset, int len); | |
7965bd4d JP |
969 | int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset, |
970 | int len); | |
971 | int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer); | |
972 | int skb_pad(struct sk_buff *skb, int pad); | |
ead2ceb0 | 973 | #define dev_kfree_skb(a) consume_skb(a) |
1da177e4 | 974 | |
7965bd4d JP |
975 | int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb, |
976 | int getfrag(void *from, char *to, int offset, | |
977 | int len, int odd, struct sk_buff *skb), | |
978 | void *from, int length); | |
e89e9cf5 | 979 | |
be12a1fe HFS |
980 | int skb_append_pagefrags(struct sk_buff *skb, struct page *page, |
981 | int offset, size_t size); | |
982 | ||
d94d9fee | 983 | struct skb_seq_state { |
677e90ed TG |
984 | __u32 lower_offset; |
985 | __u32 upper_offset; | |
986 | __u32 frag_idx; | |
987 | __u32 stepped_offset; | |
988 | struct sk_buff *root_skb; | |
989 | struct sk_buff *cur_skb; | |
990 | __u8 *frag_data; | |
991 | }; | |
992 | ||
7965bd4d JP |
993 | void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from, |
994 | unsigned int to, struct skb_seq_state *st); | |
995 | unsigned int skb_seq_read(unsigned int consumed, const u8 **data, | |
996 | struct skb_seq_state *st); | |
997 | void skb_abort_seq_read(struct skb_seq_state *st); | |
677e90ed | 998 | |
7965bd4d | 999 | unsigned int skb_find_text(struct sk_buff *skb, unsigned int from, |
059a2440 | 1000 | unsigned int to, struct ts_config *config); |
3fc7e8a6 | 1001 | |
09323cc4 TH |
1002 | /* |
1003 | * Packet hash types specify the type of hash in skb_set_hash. | |
1004 | * | |
1005 | * Hash types refer to the protocol layer addresses which are used to | |
1006 | * construct a packet's hash. The hashes are used to differentiate or identify | |
1007 | * flows of the protocol layer for the hash type. Hash types are either | |
1008 | * layer-2 (L2), layer-3 (L3), or layer-4 (L4). | |
1009 | * | |
1010 | * Properties of hashes: | |
1011 | * | |
1012 | * 1) Two packets in different flows have different hash values | |
1013 | * 2) Two packets in the same flow should have the same hash value | |
1014 | * | |
1015 | * A hash at a higher layer is considered to be more specific. A driver should | |
1016 | * set the most specific hash possible. | |
1017 | * | |
1018 | * A driver cannot indicate a more specific hash than the layer at which a hash | |
1019 | * was computed. For instance an L3 hash cannot be set as an L4 hash. | |
1020 | * | |
1021 | * A driver may indicate a hash level which is less specific than the | |
1022 | * actual layer the hash was computed on. For instance, a hash computed | |
1023 | * at L4 may be considered an L3 hash. This should only be done if the | |
1024 | * driver can't unambiguously determine that the HW computed the hash at | |
1025 | * the higher layer. Note that the "should" in the second property above | |
1026 | * permits this. | |
1027 | */ | |
1028 | enum pkt_hash_types { | |
1029 | PKT_HASH_TYPE_NONE, /* Undefined type */ | |
1030 | PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */ | |
1031 | PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */ | |
1032 | PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */ | |
1033 | }; | |
1034 | ||
bcc83839 | 1035 | static inline void skb_clear_hash(struct sk_buff *skb) |
09323cc4 | 1036 | { |
bcc83839 | 1037 | skb->hash = 0; |
a3b18ddb | 1038 | skb->sw_hash = 0; |
bcc83839 TH |
1039 | skb->l4_hash = 0; |
1040 | } | |
1041 | ||
1042 | static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb) | |
1043 | { | |
1044 | if (!skb->l4_hash) | |
1045 | skb_clear_hash(skb); | |
1046 | } | |
1047 | ||
1048 | static inline void | |
1049 | __skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4) | |
1050 | { | |
1051 | skb->l4_hash = is_l4; | |
1052 | skb->sw_hash = is_sw; | |
61b905da | 1053 | skb->hash = hash; |
09323cc4 TH |
1054 | } |
1055 | ||
bcc83839 TH |
1056 | static inline void |
1057 | skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type) | |
1058 | { | |
1059 | /* Used by drivers to set hash from HW */ | |
1060 | __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4); | |
1061 | } | |
1062 | ||
1063 | static inline void | |
1064 | __skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4) | |
1065 | { | |
1066 | __skb_set_hash(skb, hash, true, is_l4); | |
1067 | } | |
1068 | ||
e5276937 TH |
1069 | void __skb_get_hash(struct sk_buff *skb); |
1070 | u32 skb_get_poff(const struct sk_buff *skb); | |
1071 | u32 __skb_get_poff(const struct sk_buff *skb, void *data, | |
1072 | const struct flow_keys *keys, int hlen); | |
1073 | __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto, | |
1074 | void *data, int hlen_proto); | |
1075 | ||
1076 | static inline __be32 skb_flow_get_ports(const struct sk_buff *skb, | |
1077 | int thoff, u8 ip_proto) | |
1078 | { | |
1079 | return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0); | |
1080 | } | |
1081 | ||
1082 | void skb_flow_dissector_init(struct flow_dissector *flow_dissector, | |
1083 | const struct flow_dissector_key *key, | |
1084 | unsigned int key_count); | |
1085 | ||
1086 | bool __skb_flow_dissect(const struct sk_buff *skb, | |
1087 | struct flow_dissector *flow_dissector, | |
1088 | void *target_container, | |
cd79a238 TH |
1089 | void *data, __be16 proto, int nhoff, int hlen, |
1090 | unsigned int flags); | |
e5276937 TH |
1091 | |
1092 | static inline bool skb_flow_dissect(const struct sk_buff *skb, | |
1093 | struct flow_dissector *flow_dissector, | |
cd79a238 | 1094 | void *target_container, unsigned int flags) |
e5276937 TH |
1095 | { |
1096 | return __skb_flow_dissect(skb, flow_dissector, target_container, | |
cd79a238 | 1097 | NULL, 0, 0, 0, flags); |
e5276937 TH |
1098 | } |
1099 | ||
1100 | static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb, | |
cd79a238 TH |
1101 | struct flow_keys *flow, |
1102 | unsigned int flags) | |
e5276937 TH |
1103 | { |
1104 | memset(flow, 0, sizeof(*flow)); | |
1105 | return __skb_flow_dissect(skb, &flow_keys_dissector, flow, | |
cd79a238 | 1106 | NULL, 0, 0, 0, flags); |
e5276937 TH |
1107 | } |
1108 | ||
1109 | static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow, | |
1110 | void *data, __be16 proto, | |
cd79a238 TH |
1111 | int nhoff, int hlen, |
1112 | unsigned int flags) | |
e5276937 TH |
1113 | { |
1114 | memset(flow, 0, sizeof(*flow)); | |
1115 | return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow, | |
cd79a238 | 1116 | data, proto, nhoff, hlen, flags); |
e5276937 TH |
1117 | } |
1118 | ||
3958afa1 | 1119 | static inline __u32 skb_get_hash(struct sk_buff *skb) |
bfb564e7 | 1120 | { |
a3b18ddb | 1121 | if (!skb->l4_hash && !skb->sw_hash) |
3958afa1 | 1122 | __skb_get_hash(skb); |
bfb564e7 | 1123 | |
61b905da | 1124 | return skb->hash; |
bfb564e7 KK |
1125 | } |
1126 | ||
20a17bf6 | 1127 | __u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6); |
f70ea018 | 1128 | |
20a17bf6 | 1129 | static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6) |
f70ea018 | 1130 | { |
c6cc1ca7 TH |
1131 | if (!skb->l4_hash && !skb->sw_hash) { |
1132 | struct flow_keys keys; | |
de4c1f8b | 1133 | __u32 hash = __get_hash_from_flowi6(fl6, &keys); |
c6cc1ca7 | 1134 | |
de4c1f8b | 1135 | __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys)); |
c6cc1ca7 | 1136 | } |
f70ea018 TH |
1137 | |
1138 | return skb->hash; | |
1139 | } | |
1140 | ||
20a17bf6 | 1141 | __u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl); |
f70ea018 | 1142 | |
20a17bf6 | 1143 | static inline __u32 skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4) |
f70ea018 | 1144 | { |
c6cc1ca7 TH |
1145 | if (!skb->l4_hash && !skb->sw_hash) { |
1146 | struct flow_keys keys; | |
de4c1f8b | 1147 | __u32 hash = __get_hash_from_flowi4(fl4, &keys); |
c6cc1ca7 | 1148 | |
de4c1f8b | 1149 | __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys)); |
c6cc1ca7 | 1150 | } |
f70ea018 TH |
1151 | |
1152 | return skb->hash; | |
1153 | } | |
1154 | ||
50fb7992 TH |
1155 | __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb); |
1156 | ||
57bdf7f4 TH |
1157 | static inline __u32 skb_get_hash_raw(const struct sk_buff *skb) |
1158 | { | |
61b905da | 1159 | return skb->hash; |
57bdf7f4 TH |
1160 | } |
1161 | ||
3df7a74e TH |
1162 | static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from) |
1163 | { | |
61b905da | 1164 | to->hash = from->hash; |
a3b18ddb | 1165 | to->sw_hash = from->sw_hash; |
61b905da | 1166 | to->l4_hash = from->l4_hash; |
3df7a74e TH |
1167 | }; |
1168 | ||
4305b541 ACM |
1169 | #ifdef NET_SKBUFF_DATA_USES_OFFSET |
1170 | static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) | |
1171 | { | |
1172 | return skb->head + skb->end; | |
1173 | } | |
ec47ea82 AD |
1174 | |
1175 | static inline unsigned int skb_end_offset(const struct sk_buff *skb) | |
1176 | { | |
1177 | return skb->end; | |
1178 | } | |
4305b541 ACM |
1179 | #else |
1180 | static inline unsigned char *skb_end_pointer(const struct sk_buff *skb) | |
1181 | { | |
1182 | return skb->end; | |
1183 | } | |
ec47ea82 AD |
1184 | |
1185 | static inline unsigned int skb_end_offset(const struct sk_buff *skb) | |
1186 | { | |
1187 | return skb->end - skb->head; | |
1188 | } | |
4305b541 ACM |
1189 | #endif |
1190 | ||
1da177e4 | 1191 | /* Internal */ |
4305b541 | 1192 | #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB))) |
1da177e4 | 1193 | |
ac45f602 PO |
1194 | static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb) |
1195 | { | |
1196 | return &skb_shinfo(skb)->hwtstamps; | |
1197 | } | |
1198 | ||
1da177e4 LT |
1199 | /** |
1200 | * skb_queue_empty - check if a queue is empty | |
1201 | * @list: queue head | |
1202 | * | |
1203 | * Returns true if the queue is empty, false otherwise. | |
1204 | */ | |
1205 | static inline int skb_queue_empty(const struct sk_buff_head *list) | |
1206 | { | |
fd44b93c | 1207 | return list->next == (const struct sk_buff *) list; |
1da177e4 LT |
1208 | } |
1209 | ||
fc7ebb21 DM |
1210 | /** |
1211 | * skb_queue_is_last - check if skb is the last entry in the queue | |
1212 | * @list: queue head | |
1213 | * @skb: buffer | |
1214 | * | |
1215 | * Returns true if @skb is the last buffer on the list. | |
1216 | */ | |
1217 | static inline bool skb_queue_is_last(const struct sk_buff_head *list, | |
1218 | const struct sk_buff *skb) | |
1219 | { | |
fd44b93c | 1220 | return skb->next == (const struct sk_buff *) list; |
fc7ebb21 DM |
1221 | } |
1222 | ||
832d11c5 IJ |
1223 | /** |
1224 | * skb_queue_is_first - check if skb is the first entry in the queue | |
1225 | * @list: queue head | |
1226 | * @skb: buffer | |
1227 | * | |
1228 | * Returns true if @skb is the first buffer on the list. | |
1229 | */ | |
1230 | static inline bool skb_queue_is_first(const struct sk_buff_head *list, | |
1231 | const struct sk_buff *skb) | |
1232 | { | |
fd44b93c | 1233 | return skb->prev == (const struct sk_buff *) list; |
832d11c5 IJ |
1234 | } |
1235 | ||
249c8b42 DM |
1236 | /** |
1237 | * skb_queue_next - return the next packet in the queue | |
1238 | * @list: queue head | |
1239 | * @skb: current buffer | |
1240 | * | |
1241 | * Return the next packet in @list after @skb. It is only valid to | |
1242 | * call this if skb_queue_is_last() evaluates to false. | |
1243 | */ | |
1244 | static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list, | |
1245 | const struct sk_buff *skb) | |
1246 | { | |
1247 | /* This BUG_ON may seem severe, but if we just return then we | |
1248 | * are going to dereference garbage. | |
1249 | */ | |
1250 | BUG_ON(skb_queue_is_last(list, skb)); | |
1251 | return skb->next; | |
1252 | } | |
1253 | ||
832d11c5 IJ |
1254 | /** |
1255 | * skb_queue_prev - return the prev packet in the queue | |
1256 | * @list: queue head | |
1257 | * @skb: current buffer | |
1258 | * | |
1259 | * Return the prev packet in @list before @skb. It is only valid to | |
1260 | * call this if skb_queue_is_first() evaluates to false. | |
1261 | */ | |
1262 | static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list, | |
1263 | const struct sk_buff *skb) | |
1264 | { | |
1265 | /* This BUG_ON may seem severe, but if we just return then we | |
1266 | * are going to dereference garbage. | |
1267 | */ | |
1268 | BUG_ON(skb_queue_is_first(list, skb)); | |
1269 | return skb->prev; | |
1270 | } | |
1271 | ||
1da177e4 LT |
1272 | /** |
1273 | * skb_get - reference buffer | |
1274 | * @skb: buffer to reference | |
1275 | * | |
1276 | * Makes another reference to a socket buffer and returns a pointer | |
1277 | * to the buffer. | |
1278 | */ | |
1279 | static inline struct sk_buff *skb_get(struct sk_buff *skb) | |
1280 | { | |
1281 | atomic_inc(&skb->users); | |
1282 | return skb; | |
1283 | } | |
1284 | ||
1285 | /* | |
1286 | * If users == 1, we are the only owner and are can avoid redundant | |
1287 | * atomic change. | |
1288 | */ | |
1289 | ||
1da177e4 LT |
1290 | /** |
1291 | * skb_cloned - is the buffer a clone | |
1292 | * @skb: buffer to check | |
1293 | * | |
1294 | * Returns true if the buffer was generated with skb_clone() and is | |
1295 | * one of multiple shared copies of the buffer. Cloned buffers are | |
1296 | * shared data so must not be written to under normal circumstances. | |
1297 | */ | |
1298 | static inline int skb_cloned(const struct sk_buff *skb) | |
1299 | { | |
1300 | return skb->cloned && | |
1301 | (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1; | |
1302 | } | |
1303 | ||
14bbd6a5 PS |
1304 | static inline int skb_unclone(struct sk_buff *skb, gfp_t pri) |
1305 | { | |
d0164adc | 1306 | might_sleep_if(gfpflags_allow_blocking(pri)); |
14bbd6a5 PS |
1307 | |
1308 | if (skb_cloned(skb)) | |
1309 | return pskb_expand_head(skb, 0, 0, pri); | |
1310 | ||
1311 | return 0; | |
1312 | } | |
1313 | ||
1da177e4 LT |
1314 | /** |
1315 | * skb_header_cloned - is the header a clone | |
1316 | * @skb: buffer to check | |
1317 | * | |
1318 | * Returns true if modifying the header part of the buffer requires | |
1319 | * the data to be copied. | |
1320 | */ | |
1321 | static inline int skb_header_cloned(const struct sk_buff *skb) | |
1322 | { | |
1323 | int dataref; | |
1324 | ||
1325 | if (!skb->cloned) | |
1326 | return 0; | |
1327 | ||
1328 | dataref = atomic_read(&skb_shinfo(skb)->dataref); | |
1329 | dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT); | |
1330 | return dataref != 1; | |
1331 | } | |
1332 | ||
9580bf2e ED |
1333 | static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri) |
1334 | { | |
1335 | might_sleep_if(gfpflags_allow_blocking(pri)); | |
1336 | ||
1337 | if (skb_header_cloned(skb)) | |
1338 | return pskb_expand_head(skb, 0, 0, pri); | |
1339 | ||
1340 | return 0; | |
1341 | } | |
1342 | ||
1da177e4 LT |
1343 | /** |
1344 | * skb_header_release - release reference to header | |
1345 | * @skb: buffer to operate on | |
1346 | * | |
1347 | * Drop a reference to the header part of the buffer. This is done | |
1348 | * by acquiring a payload reference. You must not read from the header | |
1349 | * part of skb->data after this. | |
f4a775d1 | 1350 | * Note : Check if you can use __skb_header_release() instead. |
1da177e4 LT |
1351 | */ |
1352 | static inline void skb_header_release(struct sk_buff *skb) | |
1353 | { | |
1354 | BUG_ON(skb->nohdr); | |
1355 | skb->nohdr = 1; | |
1356 | atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref); | |
1357 | } | |
1358 | ||
f4a775d1 ED |
1359 | /** |
1360 | * __skb_header_release - release reference to header | |
1361 | * @skb: buffer to operate on | |
1362 | * | |
1363 | * Variant of skb_header_release() assuming skb is private to caller. | |
1364 | * We can avoid one atomic operation. | |
1365 | */ | |
1366 | static inline void __skb_header_release(struct sk_buff *skb) | |
1367 | { | |
1368 | skb->nohdr = 1; | |
1369 | atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT)); | |
1370 | } | |
1371 | ||
1372 | ||
1da177e4 LT |
1373 | /** |
1374 | * skb_shared - is the buffer shared | |
1375 | * @skb: buffer to check | |
1376 | * | |
1377 | * Returns true if more than one person has a reference to this | |
1378 | * buffer. | |
1379 | */ | |
1380 | static inline int skb_shared(const struct sk_buff *skb) | |
1381 | { | |
1382 | return atomic_read(&skb->users) != 1; | |
1383 | } | |
1384 | ||
1385 | /** | |
1386 | * skb_share_check - check if buffer is shared and if so clone it | |
1387 | * @skb: buffer to check | |
1388 | * @pri: priority for memory allocation | |
1389 | * | |
1390 | * If the buffer is shared the buffer is cloned and the old copy | |
1391 | * drops a reference. A new clone with a single reference is returned. | |
1392 | * If the buffer is not shared the original buffer is returned. When | |
1393 | * being called from interrupt status or with spinlocks held pri must | |
1394 | * be GFP_ATOMIC. | |
1395 | * | |
1396 | * NULL is returned on a memory allocation failure. | |
1397 | */ | |
47061bc4 | 1398 | static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri) |
1da177e4 | 1399 | { |
d0164adc | 1400 | might_sleep_if(gfpflags_allow_blocking(pri)); |
1da177e4 LT |
1401 | if (skb_shared(skb)) { |
1402 | struct sk_buff *nskb = skb_clone(skb, pri); | |
47061bc4 ED |
1403 | |
1404 | if (likely(nskb)) | |
1405 | consume_skb(skb); | |
1406 | else | |
1407 | kfree_skb(skb); | |
1da177e4 LT |
1408 | skb = nskb; |
1409 | } | |
1410 | return skb; | |
1411 | } | |
1412 | ||
1413 | /* | |
1414 | * Copy shared buffers into a new sk_buff. We effectively do COW on | |
1415 | * packets to handle cases where we have a local reader and forward | |
1416 | * and a couple of other messy ones. The normal one is tcpdumping | |
1417 | * a packet thats being forwarded. | |
1418 | */ | |
1419 | ||
1420 | /** | |
1421 | * skb_unshare - make a copy of a shared buffer | |
1422 | * @skb: buffer to check | |
1423 | * @pri: priority for memory allocation | |
1424 | * | |
1425 | * If the socket buffer is a clone then this function creates a new | |
1426 | * copy of the data, drops a reference count on the old copy and returns | |
1427 | * the new copy with the reference count at 1. If the buffer is not a clone | |
1428 | * the original buffer is returned. When called with a spinlock held or | |
1429 | * from interrupt state @pri must be %GFP_ATOMIC | |
1430 | * | |
1431 | * %NULL is returned on a memory allocation failure. | |
1432 | */ | |
e2bf521d | 1433 | static inline struct sk_buff *skb_unshare(struct sk_buff *skb, |
dd0fc66f | 1434 | gfp_t pri) |
1da177e4 | 1435 | { |
d0164adc | 1436 | might_sleep_if(gfpflags_allow_blocking(pri)); |
1da177e4 LT |
1437 | if (skb_cloned(skb)) { |
1438 | struct sk_buff *nskb = skb_copy(skb, pri); | |
31eff81e AA |
1439 | |
1440 | /* Free our shared copy */ | |
1441 | if (likely(nskb)) | |
1442 | consume_skb(skb); | |
1443 | else | |
1444 | kfree_skb(skb); | |
1da177e4 LT |
1445 | skb = nskb; |
1446 | } | |
1447 | return skb; | |
1448 | } | |
1449 | ||
1450 | /** | |
1a5778aa | 1451 | * skb_peek - peek at the head of an &sk_buff_head |
1da177e4 LT |
1452 | * @list_: list to peek at |
1453 | * | |
1454 | * Peek an &sk_buff. Unlike most other operations you _MUST_ | |
1455 | * be careful with this one. A peek leaves the buffer on the | |
1456 | * list and someone else may run off with it. You must hold | |
1457 | * the appropriate locks or have a private queue to do this. | |
1458 | * | |
1459 | * Returns %NULL for an empty list or a pointer to the head element. | |
1460 | * The reference count is not incremented and the reference is therefore | |
1461 | * volatile. Use with caution. | |
1462 | */ | |
05bdd2f1 | 1463 | static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_) |
1da177e4 | 1464 | { |
18d07000 ED |
1465 | struct sk_buff *skb = list_->next; |
1466 | ||
1467 | if (skb == (struct sk_buff *)list_) | |
1468 | skb = NULL; | |
1469 | return skb; | |
1da177e4 LT |
1470 | } |
1471 | ||
da5ef6e5 PE |
1472 | /** |
1473 | * skb_peek_next - peek skb following the given one from a queue | |
1474 | * @skb: skb to start from | |
1475 | * @list_: list to peek at | |
1476 | * | |
1477 | * Returns %NULL when the end of the list is met or a pointer to the | |
1478 | * next element. The reference count is not incremented and the | |
1479 | * reference is therefore volatile. Use with caution. | |
1480 | */ | |
1481 | static inline struct sk_buff *skb_peek_next(struct sk_buff *skb, | |
1482 | const struct sk_buff_head *list_) | |
1483 | { | |
1484 | struct sk_buff *next = skb->next; | |
18d07000 | 1485 | |
da5ef6e5 PE |
1486 | if (next == (struct sk_buff *)list_) |
1487 | next = NULL; | |
1488 | return next; | |
1489 | } | |
1490 | ||
1da177e4 | 1491 | /** |
1a5778aa | 1492 | * skb_peek_tail - peek at the tail of an &sk_buff_head |
1da177e4 LT |
1493 | * @list_: list to peek at |
1494 | * | |
1495 | * Peek an &sk_buff. Unlike most other operations you _MUST_ | |
1496 | * be careful with this one. A peek leaves the buffer on the | |
1497 | * list and someone else may run off with it. You must hold | |
1498 | * the appropriate locks or have a private queue to do this. | |
1499 | * | |
1500 | * Returns %NULL for an empty list or a pointer to the tail element. | |
1501 | * The reference count is not incremented and the reference is therefore | |
1502 | * volatile. Use with caution. | |
1503 | */ | |
05bdd2f1 | 1504 | static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_) |
1da177e4 | 1505 | { |
18d07000 ED |
1506 | struct sk_buff *skb = list_->prev; |
1507 | ||
1508 | if (skb == (struct sk_buff *)list_) | |
1509 | skb = NULL; | |
1510 | return skb; | |
1511 | ||
1da177e4 LT |
1512 | } |
1513 | ||
1514 | /** | |
1515 | * skb_queue_len - get queue length | |
1516 | * @list_: list to measure | |
1517 | * | |
1518 | * Return the length of an &sk_buff queue. | |
1519 | */ | |
1520 | static inline __u32 skb_queue_len(const struct sk_buff_head *list_) | |
1521 | { | |
1522 | return list_->qlen; | |
1523 | } | |
1524 | ||
67fed459 DM |
1525 | /** |
1526 | * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head | |
1527 | * @list: queue to initialize | |
1528 | * | |
1529 | * This initializes only the list and queue length aspects of | |
1530 | * an sk_buff_head object. This allows to initialize the list | |
1531 | * aspects of an sk_buff_head without reinitializing things like | |
1532 | * the spinlock. It can also be used for on-stack sk_buff_head | |
1533 | * objects where the spinlock is known to not be used. | |
1534 | */ | |
1535 | static inline void __skb_queue_head_init(struct sk_buff_head *list) | |
1536 | { | |
1537 | list->prev = list->next = (struct sk_buff *)list; | |
1538 | list->qlen = 0; | |
1539 | } | |
1540 | ||
76f10ad0 AV |
1541 | /* |
1542 | * This function creates a split out lock class for each invocation; | |
1543 | * this is needed for now since a whole lot of users of the skb-queue | |
1544 | * infrastructure in drivers have different locking usage (in hardirq) | |
1545 | * than the networking core (in softirq only). In the long run either the | |
1546 | * network layer or drivers should need annotation to consolidate the | |
1547 | * main types of usage into 3 classes. | |
1548 | */ | |
1da177e4 LT |
1549 | static inline void skb_queue_head_init(struct sk_buff_head *list) |
1550 | { | |
1551 | spin_lock_init(&list->lock); | |
67fed459 | 1552 | __skb_queue_head_init(list); |
1da177e4 LT |
1553 | } |
1554 | ||
c2ecba71 PE |
1555 | static inline void skb_queue_head_init_class(struct sk_buff_head *list, |
1556 | struct lock_class_key *class) | |
1557 | { | |
1558 | skb_queue_head_init(list); | |
1559 | lockdep_set_class(&list->lock, class); | |
1560 | } | |
1561 | ||
1da177e4 | 1562 | /* |
bf299275 | 1563 | * Insert an sk_buff on a list. |
1da177e4 LT |
1564 | * |
1565 | * The "__skb_xxxx()" functions are the non-atomic ones that | |
1566 | * can only be called with interrupts disabled. | |
1567 | */ | |
7965bd4d JP |
1568 | void skb_insert(struct sk_buff *old, struct sk_buff *newsk, |
1569 | struct sk_buff_head *list); | |
bf299275 GR |
1570 | static inline void __skb_insert(struct sk_buff *newsk, |
1571 | struct sk_buff *prev, struct sk_buff *next, | |
1572 | struct sk_buff_head *list) | |
1573 | { | |
1574 | newsk->next = next; | |
1575 | newsk->prev = prev; | |
1576 | next->prev = prev->next = newsk; | |
1577 | list->qlen++; | |
1578 | } | |
1da177e4 | 1579 | |
67fed459 DM |
1580 | static inline void __skb_queue_splice(const struct sk_buff_head *list, |
1581 | struct sk_buff *prev, | |
1582 | struct sk_buff *next) | |
1583 | { | |
1584 | struct sk_buff *first = list->next; | |
1585 | struct sk_buff *last = list->prev; | |
1586 | ||
1587 | first->prev = prev; | |
1588 | prev->next = first; | |
1589 | ||
1590 | last->next = next; | |
1591 | next->prev = last; | |
1592 | } | |
1593 | ||
1594 | /** | |
1595 | * skb_queue_splice - join two skb lists, this is designed for stacks | |
1596 | * @list: the new list to add | |
1597 | * @head: the place to add it in the first list | |
1598 | */ | |
1599 | static inline void skb_queue_splice(const struct sk_buff_head *list, | |
1600 | struct sk_buff_head *head) | |
1601 | { | |
1602 | if (!skb_queue_empty(list)) { | |
1603 | __skb_queue_splice(list, (struct sk_buff *) head, head->next); | |
1d4a31dd | 1604 | head->qlen += list->qlen; |
67fed459 DM |
1605 | } |
1606 | } | |
1607 | ||
1608 | /** | |
d9619496 | 1609 | * skb_queue_splice_init - join two skb lists and reinitialise the emptied list |
67fed459 DM |
1610 | * @list: the new list to add |
1611 | * @head: the place to add it in the first list | |
1612 | * | |
1613 | * The list at @list is reinitialised | |
1614 | */ | |
1615 | static inline void skb_queue_splice_init(struct sk_buff_head *list, | |
1616 | struct sk_buff_head *head) | |
1617 | { | |
1618 | if (!skb_queue_empty(list)) { | |
1619 | __skb_queue_splice(list, (struct sk_buff *) head, head->next); | |
1d4a31dd | 1620 | head->qlen += list->qlen; |
67fed459 DM |
1621 | __skb_queue_head_init(list); |
1622 | } | |
1623 | } | |
1624 | ||
1625 | /** | |
1626 | * skb_queue_splice_tail - join two skb lists, each list being a queue | |
1627 | * @list: the new list to add | |
1628 | * @head: the place to add it in the first list | |
1629 | */ | |
1630 | static inline void skb_queue_splice_tail(const struct sk_buff_head *list, | |
1631 | struct sk_buff_head *head) | |
1632 | { | |
1633 | if (!skb_queue_empty(list)) { | |
1634 | __skb_queue_splice(list, head->prev, (struct sk_buff *) head); | |
1d4a31dd | 1635 | head->qlen += list->qlen; |
67fed459 DM |
1636 | } |
1637 | } | |
1638 | ||
1639 | /** | |
d9619496 | 1640 | * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list |
67fed459 DM |
1641 | * @list: the new list to add |
1642 | * @head: the place to add it in the first list | |
1643 | * | |
1644 | * Each of the lists is a queue. | |
1645 | * The list at @list is reinitialised | |
1646 | */ | |
1647 | static inline void skb_queue_splice_tail_init(struct sk_buff_head *list, | |
1648 | struct sk_buff_head *head) | |
1649 | { | |
1650 | if (!skb_queue_empty(list)) { | |
1651 | __skb_queue_splice(list, head->prev, (struct sk_buff *) head); | |
1d4a31dd | 1652 | head->qlen += list->qlen; |
67fed459 DM |
1653 | __skb_queue_head_init(list); |
1654 | } | |
1655 | } | |
1656 | ||
1da177e4 | 1657 | /** |
300ce174 | 1658 | * __skb_queue_after - queue a buffer at the list head |
1da177e4 | 1659 | * @list: list to use |
300ce174 | 1660 | * @prev: place after this buffer |
1da177e4 LT |
1661 | * @newsk: buffer to queue |
1662 | * | |
300ce174 | 1663 | * Queue a buffer int the middle of a list. This function takes no locks |
1da177e4 LT |
1664 | * and you must therefore hold required locks before calling it. |
1665 | * | |
1666 | * A buffer cannot be placed on two lists at the same time. | |
1667 | */ | |
300ce174 SH |
1668 | static inline void __skb_queue_after(struct sk_buff_head *list, |
1669 | struct sk_buff *prev, | |
1670 | struct sk_buff *newsk) | |
1da177e4 | 1671 | { |
bf299275 | 1672 | __skb_insert(newsk, prev, prev->next, list); |
1da177e4 LT |
1673 | } |
1674 | ||
7965bd4d JP |
1675 | void skb_append(struct sk_buff *old, struct sk_buff *newsk, |
1676 | struct sk_buff_head *list); | |
7de6c033 | 1677 | |
f5572855 GR |
1678 | static inline void __skb_queue_before(struct sk_buff_head *list, |
1679 | struct sk_buff *next, | |
1680 | struct sk_buff *newsk) | |
1681 | { | |
1682 | __skb_insert(newsk, next->prev, next, list); | |
1683 | } | |
1684 | ||
300ce174 SH |
1685 | /** |
1686 | * __skb_queue_head - queue a buffer at the list head | |
1687 | * @list: list to use | |
1688 | * @newsk: buffer to queue | |
1689 | * | |
1690 | * Queue a buffer at the start of a list. This function takes no locks | |
1691 | * and you must therefore hold required locks before calling it. | |
1692 | * | |
1693 | * A buffer cannot be placed on two lists at the same time. | |
1694 | */ | |
7965bd4d | 1695 | void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk); |
300ce174 SH |
1696 | static inline void __skb_queue_head(struct sk_buff_head *list, |
1697 | struct sk_buff *newsk) | |
1698 | { | |
1699 | __skb_queue_after(list, (struct sk_buff *)list, newsk); | |
1700 | } | |
1701 | ||
1da177e4 LT |
1702 | /** |
1703 | * __skb_queue_tail - queue a buffer at the list tail | |
1704 | * @list: list to use | |
1705 | * @newsk: buffer to queue | |
1706 | * | |
1707 | * Queue a buffer at the end of a list. This function takes no locks | |
1708 | * and you must therefore hold required locks before calling it. | |
1709 | * | |
1710 | * A buffer cannot be placed on two lists at the same time. | |
1711 | */ | |
7965bd4d | 1712 | void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk); |
1da177e4 LT |
1713 | static inline void __skb_queue_tail(struct sk_buff_head *list, |
1714 | struct sk_buff *newsk) | |
1715 | { | |
f5572855 | 1716 | __skb_queue_before(list, (struct sk_buff *)list, newsk); |
1da177e4 LT |
1717 | } |
1718 | ||
1da177e4 LT |
1719 | /* |
1720 | * remove sk_buff from list. _Must_ be called atomically, and with | |
1721 | * the list known.. | |
1722 | */ | |
7965bd4d | 1723 | void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list); |
1da177e4 LT |
1724 | static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list) |
1725 | { | |
1726 | struct sk_buff *next, *prev; | |
1727 | ||
1728 | list->qlen--; | |
1729 | next = skb->next; | |
1730 | prev = skb->prev; | |
1731 | skb->next = skb->prev = NULL; | |
1da177e4 LT |
1732 | next->prev = prev; |
1733 | prev->next = next; | |
1734 | } | |
1735 | ||
f525c06d GR |
1736 | /** |
1737 | * __skb_dequeue - remove from the head of the queue | |
1738 | * @list: list to dequeue from | |
1739 | * | |
1740 | * Remove the head of the list. This function does not take any locks | |
1741 | * so must be used with appropriate locks held only. The head item is | |
1742 | * returned or %NULL if the list is empty. | |
1743 | */ | |
7965bd4d | 1744 | struct sk_buff *skb_dequeue(struct sk_buff_head *list); |
f525c06d GR |
1745 | static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list) |
1746 | { | |
1747 | struct sk_buff *skb = skb_peek(list); | |
1748 | if (skb) | |
1749 | __skb_unlink(skb, list); | |
1750 | return skb; | |
1751 | } | |
1da177e4 LT |
1752 | |
1753 | /** | |
1754 | * __skb_dequeue_tail - remove from the tail of the queue | |
1755 | * @list: list to dequeue from | |
1756 | * | |
1757 | * Remove the tail of the list. This function does not take any locks | |
1758 | * so must be used with appropriate locks held only. The tail item is | |
1759 | * returned or %NULL if the list is empty. | |
1760 | */ | |
7965bd4d | 1761 | struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list); |
1da177e4 LT |
1762 | static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list) |
1763 | { | |
1764 | struct sk_buff *skb = skb_peek_tail(list); | |
1765 | if (skb) | |
1766 | __skb_unlink(skb, list); | |
1767 | return skb; | |
1768 | } | |
1769 | ||
1770 | ||
bdcc0924 | 1771 | static inline bool skb_is_nonlinear(const struct sk_buff *skb) |
1da177e4 LT |
1772 | { |
1773 | return skb->data_len; | |
1774 | } | |
1775 | ||
1776 | static inline unsigned int skb_headlen(const struct sk_buff *skb) | |
1777 | { | |
1778 | return skb->len - skb->data_len; | |
1779 | } | |
1780 | ||
1781 | static inline int skb_pagelen(const struct sk_buff *skb) | |
1782 | { | |
1783 | int i, len = 0; | |
1784 | ||
1785 | for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--) | |
9e903e08 | 1786 | len += skb_frag_size(&skb_shinfo(skb)->frags[i]); |
1da177e4 LT |
1787 | return len + skb_headlen(skb); |
1788 | } | |
1789 | ||
131ea667 IC |
1790 | /** |
1791 | * __skb_fill_page_desc - initialise a paged fragment in an skb | |
1792 | * @skb: buffer containing fragment to be initialised | |
1793 | * @i: paged fragment index to initialise | |
1794 | * @page: the page to use for this fragment | |
1795 | * @off: the offset to the data with @page | |
1796 | * @size: the length of the data | |
1797 | * | |
1798 | * Initialises the @i'th fragment of @skb to point to &size bytes at | |
1799 | * offset @off within @page. | |
1800 | * | |
1801 | * Does not take any additional reference on the fragment. | |
1802 | */ | |
1803 | static inline void __skb_fill_page_desc(struct sk_buff *skb, int i, | |
1804 | struct page *page, int off, int size) | |
1da177e4 LT |
1805 | { |
1806 | skb_frag_t *frag = &skb_shinfo(skb)->frags[i]; | |
1807 | ||
c48a11c7 | 1808 | /* |
2f064f34 MH |
1809 | * Propagate page pfmemalloc to the skb if we can. The problem is |
1810 | * that not all callers have unique ownership of the page but rely | |
1811 | * on page_is_pfmemalloc doing the right thing(tm). | |
c48a11c7 | 1812 | */ |
a8605c60 | 1813 | frag->page.p = page; |
1da177e4 | 1814 | frag->page_offset = off; |
9e903e08 | 1815 | skb_frag_size_set(frag, size); |
cca7af38 PE |
1816 | |
1817 | page = compound_head(page); | |
2f064f34 | 1818 | if (page_is_pfmemalloc(page)) |
cca7af38 | 1819 | skb->pfmemalloc = true; |
131ea667 IC |
1820 | } |
1821 | ||
1822 | /** | |
1823 | * skb_fill_page_desc - initialise a paged fragment in an skb | |
1824 | * @skb: buffer containing fragment to be initialised | |
1825 | * @i: paged fragment index to initialise | |
1826 | * @page: the page to use for this fragment | |
1827 | * @off: the offset to the data with @page | |
1828 | * @size: the length of the data | |
1829 | * | |
1830 | * As per __skb_fill_page_desc() -- initialises the @i'th fragment of | |
bc32383c | 1831 | * @skb to point to @size bytes at offset @off within @page. In |
131ea667 IC |
1832 | * addition updates @skb such that @i is the last fragment. |
1833 | * | |
1834 | * Does not take any additional reference on the fragment. | |
1835 | */ | |
1836 | static inline void skb_fill_page_desc(struct sk_buff *skb, int i, | |
1837 | struct page *page, int off, int size) | |
1838 | { | |
1839 | __skb_fill_page_desc(skb, i, page, off, size); | |
1da177e4 LT |
1840 | skb_shinfo(skb)->nr_frags = i + 1; |
1841 | } | |
1842 | ||
7965bd4d JP |
1843 | void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off, |
1844 | int size, unsigned int truesize); | |
654bed16 | 1845 | |
f8e617e1 JW |
1846 | void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size, |
1847 | unsigned int truesize); | |
1848 | ||
1da177e4 | 1849 | #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags) |
21dc3301 | 1850 | #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb)) |
1da177e4 LT |
1851 | #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb)) |
1852 | ||
27a884dc ACM |
1853 | #ifdef NET_SKBUFF_DATA_USES_OFFSET |
1854 | static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) | |
1855 | { | |
1856 | return skb->head + skb->tail; | |
1857 | } | |
1858 | ||
1859 | static inline void skb_reset_tail_pointer(struct sk_buff *skb) | |
1860 | { | |
1861 | skb->tail = skb->data - skb->head; | |
1862 | } | |
1863 | ||
1864 | static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) | |
1865 | { | |
1866 | skb_reset_tail_pointer(skb); | |
1867 | skb->tail += offset; | |
1868 | } | |
7cc46190 | 1869 | |
27a884dc ACM |
1870 | #else /* NET_SKBUFF_DATA_USES_OFFSET */ |
1871 | static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb) | |
1872 | { | |
1873 | return skb->tail; | |
1874 | } | |
1875 | ||
1876 | static inline void skb_reset_tail_pointer(struct sk_buff *skb) | |
1877 | { | |
1878 | skb->tail = skb->data; | |
1879 | } | |
1880 | ||
1881 | static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset) | |
1882 | { | |
1883 | skb->tail = skb->data + offset; | |
1884 | } | |
4305b541 | 1885 | |
27a884dc ACM |
1886 | #endif /* NET_SKBUFF_DATA_USES_OFFSET */ |
1887 | ||
1da177e4 LT |
1888 | /* |
1889 | * Add data to an sk_buff | |
1890 | */ | |
0c7ddf36 | 1891 | unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len); |
7965bd4d | 1892 | unsigned char *skb_put(struct sk_buff *skb, unsigned int len); |
1da177e4 LT |
1893 | static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len) |
1894 | { | |
27a884dc | 1895 | unsigned char *tmp = skb_tail_pointer(skb); |
1da177e4 LT |
1896 | SKB_LINEAR_ASSERT(skb); |
1897 | skb->tail += len; | |
1898 | skb->len += len; | |
1899 | return tmp; | |
1900 | } | |
1901 | ||
7965bd4d | 1902 | unsigned char *skb_push(struct sk_buff *skb, unsigned int len); |
1da177e4 LT |
1903 | static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len) |
1904 | { | |
1905 | skb->data -= len; | |
1906 | skb->len += len; | |
1907 | return skb->data; | |
1908 | } | |
1909 | ||
7965bd4d | 1910 | unsigned char *skb_pull(struct sk_buff *skb, unsigned int len); |
1da177e4 LT |
1911 | static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len) |
1912 | { | |
1913 | skb->len -= len; | |
1914 | BUG_ON(skb->len < skb->data_len); | |
1915 | return skb->data += len; | |
1916 | } | |
1917 | ||
47d29646 DM |
1918 | static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len) |
1919 | { | |
1920 | return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len); | |
1921 | } | |
1922 | ||
7965bd4d | 1923 | unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta); |
1da177e4 LT |
1924 | |
1925 | static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len) | |
1926 | { | |
1927 | if (len > skb_headlen(skb) && | |
987c402a | 1928 | !__pskb_pull_tail(skb, len - skb_headlen(skb))) |
1da177e4 LT |
1929 | return NULL; |
1930 | skb->len -= len; | |
1931 | return skb->data += len; | |
1932 | } | |
1933 | ||
1934 | static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len) | |
1935 | { | |
1936 | return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len); | |
1937 | } | |
1938 | ||
1939 | static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len) | |
1940 | { | |
1941 | if (likely(len <= skb_headlen(skb))) | |
1942 | return 1; | |
1943 | if (unlikely(len > skb->len)) | |
1944 | return 0; | |
987c402a | 1945 | return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL; |
1da177e4 LT |
1946 | } |
1947 | ||
1948 | /** | |
1949 | * skb_headroom - bytes at buffer head | |
1950 | * @skb: buffer to check | |
1951 | * | |
1952 | * Return the number of bytes of free space at the head of an &sk_buff. | |
1953 | */ | |
c2636b4d | 1954 | static inline unsigned int skb_headroom(const struct sk_buff *skb) |
1da177e4 LT |
1955 | { |
1956 | return skb->data - skb->head; | |
1957 | } | |
1958 | ||
1959 | /** | |
1960 | * skb_tailroom - bytes at buffer end | |
1961 | * @skb: buffer to check | |
1962 | * | |
1963 | * Return the number of bytes of free space at the tail of an sk_buff | |
1964 | */ | |
1965 | static inline int skb_tailroom(const struct sk_buff *skb) | |
1966 | { | |
4305b541 | 1967 | return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail; |
1da177e4 LT |
1968 | } |
1969 | ||
a21d4572 ED |
1970 | /** |
1971 | * skb_availroom - bytes at buffer end | |
1972 | * @skb: buffer to check | |
1973 | * | |
1974 | * Return the number of bytes of free space at the tail of an sk_buff | |
1975 | * allocated by sk_stream_alloc() | |
1976 | */ | |
1977 | static inline int skb_availroom(const struct sk_buff *skb) | |
1978 | { | |
16fad69c ED |
1979 | if (skb_is_nonlinear(skb)) |
1980 | return 0; | |
1981 | ||
1982 | return skb->end - skb->tail - skb->reserved_tailroom; | |
a21d4572 ED |
1983 | } |
1984 | ||
1da177e4 LT |
1985 | /** |
1986 | * skb_reserve - adjust headroom | |
1987 | * @skb: buffer to alter | |
1988 | * @len: bytes to move | |
1989 | * | |
1990 | * Increase the headroom of an empty &sk_buff by reducing the tail | |
1991 | * room. This is only allowed for an empty buffer. | |
1992 | */ | |
8243126c | 1993 | static inline void skb_reserve(struct sk_buff *skb, int len) |
1da177e4 LT |
1994 | { |
1995 | skb->data += len; | |
1996 | skb->tail += len; | |
1997 | } | |
1998 | ||
1837b2e2 BP |
1999 | /** |
2000 | * skb_tailroom_reserve - adjust reserved_tailroom | |
2001 | * @skb: buffer to alter | |
2002 | * @mtu: maximum amount of headlen permitted | |
2003 | * @needed_tailroom: minimum amount of reserved_tailroom | |
2004 | * | |
2005 | * Set reserved_tailroom so that headlen can be as large as possible but | |
2006 | * not larger than mtu and tailroom cannot be smaller than | |
2007 | * needed_tailroom. | |
2008 | * The required headroom should already have been reserved before using | |
2009 | * this function. | |
2010 | */ | |
2011 | static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu, | |
2012 | unsigned int needed_tailroom) | |
2013 | { | |
2014 | SKB_LINEAR_ASSERT(skb); | |
2015 | if (mtu < skb_tailroom(skb) - needed_tailroom) | |
2016 | /* use at most mtu */ | |
2017 | skb->reserved_tailroom = skb_tailroom(skb) - mtu; | |
2018 | else | |
2019 | /* use up to all available space */ | |
2020 | skb->reserved_tailroom = needed_tailroom; | |
2021 | } | |
2022 | ||
8bce6d7d TH |
2023 | #define ENCAP_TYPE_ETHER 0 |
2024 | #define ENCAP_TYPE_IPPROTO 1 | |
2025 | ||
2026 | static inline void skb_set_inner_protocol(struct sk_buff *skb, | |
2027 | __be16 protocol) | |
2028 | { | |
2029 | skb->inner_protocol = protocol; | |
2030 | skb->inner_protocol_type = ENCAP_TYPE_ETHER; | |
2031 | } | |
2032 | ||
2033 | static inline void skb_set_inner_ipproto(struct sk_buff *skb, | |
2034 | __u8 ipproto) | |
2035 | { | |
2036 | skb->inner_ipproto = ipproto; | |
2037 | skb->inner_protocol_type = ENCAP_TYPE_IPPROTO; | |
2038 | } | |
2039 | ||
6a674e9c JG |
2040 | static inline void skb_reset_inner_headers(struct sk_buff *skb) |
2041 | { | |
aefbd2b3 | 2042 | skb->inner_mac_header = skb->mac_header; |
6a674e9c JG |
2043 | skb->inner_network_header = skb->network_header; |
2044 | skb->inner_transport_header = skb->transport_header; | |
2045 | } | |
2046 | ||
0b5c9db1 JP |
2047 | static inline void skb_reset_mac_len(struct sk_buff *skb) |
2048 | { | |
2049 | skb->mac_len = skb->network_header - skb->mac_header; | |
2050 | } | |
2051 | ||
6a674e9c JG |
2052 | static inline unsigned char *skb_inner_transport_header(const struct sk_buff |
2053 | *skb) | |
2054 | { | |
2055 | return skb->head + skb->inner_transport_header; | |
2056 | } | |
2057 | ||
55dc5a9f TH |
2058 | static inline int skb_inner_transport_offset(const struct sk_buff *skb) |
2059 | { | |
2060 | return skb_inner_transport_header(skb) - skb->data; | |
2061 | } | |
2062 | ||
6a674e9c JG |
2063 | static inline void skb_reset_inner_transport_header(struct sk_buff *skb) |
2064 | { | |
2065 | skb->inner_transport_header = skb->data - skb->head; | |
2066 | } | |
2067 | ||
2068 | static inline void skb_set_inner_transport_header(struct sk_buff *skb, | |
2069 | const int offset) | |
2070 | { | |
2071 | skb_reset_inner_transport_header(skb); | |
2072 | skb->inner_transport_header += offset; | |
2073 | } | |
2074 | ||
2075 | static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb) | |
2076 | { | |
2077 | return skb->head + skb->inner_network_header; | |
2078 | } | |
2079 | ||
2080 | static inline void skb_reset_inner_network_header(struct sk_buff *skb) | |
2081 | { | |
2082 | skb->inner_network_header = skb->data - skb->head; | |
2083 | } | |
2084 | ||
2085 | static inline void skb_set_inner_network_header(struct sk_buff *skb, | |
2086 | const int offset) | |
2087 | { | |
2088 | skb_reset_inner_network_header(skb); | |
2089 | skb->inner_network_header += offset; | |
2090 | } | |
2091 | ||
aefbd2b3 PS |
2092 | static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb) |
2093 | { | |
2094 | return skb->head + skb->inner_mac_header; | |
2095 | } | |
2096 | ||
2097 | static inline void skb_reset_inner_mac_header(struct sk_buff *skb) | |
2098 | { | |
2099 | skb->inner_mac_header = skb->data - skb->head; | |
2100 | } | |
2101 | ||
2102 | static inline void skb_set_inner_mac_header(struct sk_buff *skb, | |
2103 | const int offset) | |
2104 | { | |
2105 | skb_reset_inner_mac_header(skb); | |
2106 | skb->inner_mac_header += offset; | |
2107 | } | |
fda55eca ED |
2108 | static inline bool skb_transport_header_was_set(const struct sk_buff *skb) |
2109 | { | |
35d04610 | 2110 | return skb->transport_header != (typeof(skb->transport_header))~0U; |
fda55eca ED |
2111 | } |
2112 | ||
9c70220b ACM |
2113 | static inline unsigned char *skb_transport_header(const struct sk_buff *skb) |
2114 | { | |
2e07fa9c | 2115 | return skb->head + skb->transport_header; |
9c70220b ACM |
2116 | } |
2117 | ||
badff6d0 ACM |
2118 | static inline void skb_reset_transport_header(struct sk_buff *skb) |
2119 | { | |
2e07fa9c | 2120 | skb->transport_header = skb->data - skb->head; |
badff6d0 ACM |
2121 | } |
2122 | ||
967b05f6 ACM |
2123 | static inline void skb_set_transport_header(struct sk_buff *skb, |
2124 | const int offset) | |
2125 | { | |
2e07fa9c ACM |
2126 | skb_reset_transport_header(skb); |
2127 | skb->transport_header += offset; | |
ea2ae17d ACM |
2128 | } |
2129 | ||
d56f90a7 ACM |
2130 | static inline unsigned char *skb_network_header(const struct sk_buff *skb) |
2131 | { | |
2e07fa9c | 2132 | return skb->head + skb->network_header; |
d56f90a7 ACM |
2133 | } |
2134 | ||
c1d2bbe1 ACM |
2135 | static inline void skb_reset_network_header(struct sk_buff *skb) |
2136 | { | |
2e07fa9c | 2137 | skb->network_header = skb->data - skb->head; |
c1d2bbe1 ACM |
2138 | } |
2139 | ||
c14d2450 ACM |
2140 | static inline void skb_set_network_header(struct sk_buff *skb, const int offset) |
2141 | { | |
2e07fa9c ACM |
2142 | skb_reset_network_header(skb); |
2143 | skb->network_header += offset; | |
c14d2450 ACM |
2144 | } |
2145 | ||
2e07fa9c | 2146 | static inline unsigned char *skb_mac_header(const struct sk_buff *skb) |
bbe735e4 | 2147 | { |
2e07fa9c | 2148 | return skb->head + skb->mac_header; |
bbe735e4 ACM |
2149 | } |
2150 | ||
2e07fa9c | 2151 | static inline int skb_mac_header_was_set(const struct sk_buff *skb) |
cfe1fc77 | 2152 | { |
35d04610 | 2153 | return skb->mac_header != (typeof(skb->mac_header))~0U; |
2e07fa9c ACM |
2154 | } |
2155 | ||
2156 | static inline void skb_reset_mac_header(struct sk_buff *skb) | |
2157 | { | |
2158 | skb->mac_header = skb->data - skb->head; | |
2159 | } | |
2160 | ||
2161 | static inline void skb_set_mac_header(struct sk_buff *skb, const int offset) | |
2162 | { | |
2163 | skb_reset_mac_header(skb); | |
2164 | skb->mac_header += offset; | |
2165 | } | |
2166 | ||
0e3da5bb TT |
2167 | static inline void skb_pop_mac_header(struct sk_buff *skb) |
2168 | { | |
2169 | skb->mac_header = skb->network_header; | |
2170 | } | |
2171 | ||
fbbdb8f0 YX |
2172 | static inline void skb_probe_transport_header(struct sk_buff *skb, |
2173 | const int offset_hint) | |
2174 | { | |
2175 | struct flow_keys keys; | |
2176 | ||
2177 | if (skb_transport_header_was_set(skb)) | |
2178 | return; | |
cd79a238 | 2179 | else if (skb_flow_dissect_flow_keys(skb, &keys, 0)) |
42aecaa9 | 2180 | skb_set_transport_header(skb, keys.control.thoff); |
fbbdb8f0 YX |
2181 | else |
2182 | skb_set_transport_header(skb, offset_hint); | |
2183 | } | |
2184 | ||
03606895 ED |
2185 | static inline void skb_mac_header_rebuild(struct sk_buff *skb) |
2186 | { | |
2187 | if (skb_mac_header_was_set(skb)) { | |
2188 | const unsigned char *old_mac = skb_mac_header(skb); | |
2189 | ||
2190 | skb_set_mac_header(skb, -skb->mac_len); | |
2191 | memmove(skb_mac_header(skb), old_mac, skb->mac_len); | |
2192 | } | |
2193 | } | |
2194 | ||
04fb451e MM |
2195 | static inline int skb_checksum_start_offset(const struct sk_buff *skb) |
2196 | { | |
2197 | return skb->csum_start - skb_headroom(skb); | |
2198 | } | |
2199 | ||
08b64fcc AD |
2200 | static inline unsigned char *skb_checksum_start(const struct sk_buff *skb) |
2201 | { | |
2202 | return skb->head + skb->csum_start; | |
2203 | } | |
2204 | ||
2e07fa9c ACM |
2205 | static inline int skb_transport_offset(const struct sk_buff *skb) |
2206 | { | |
2207 | return skb_transport_header(skb) - skb->data; | |
2208 | } | |
2209 | ||
2210 | static inline u32 skb_network_header_len(const struct sk_buff *skb) | |
2211 | { | |
2212 | return skb->transport_header - skb->network_header; | |
2213 | } | |
2214 | ||
6a674e9c JG |
2215 | static inline u32 skb_inner_network_header_len(const struct sk_buff *skb) |
2216 | { | |
2217 | return skb->inner_transport_header - skb->inner_network_header; | |
2218 | } | |
2219 | ||
2e07fa9c ACM |
2220 | static inline int skb_network_offset(const struct sk_buff *skb) |
2221 | { | |
2222 | return skb_network_header(skb) - skb->data; | |
2223 | } | |
48d49d0c | 2224 | |
6a674e9c JG |
2225 | static inline int skb_inner_network_offset(const struct sk_buff *skb) |
2226 | { | |
2227 | return skb_inner_network_header(skb) - skb->data; | |
2228 | } | |
2229 | ||
f9599ce1 CG |
2230 | static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len) |
2231 | { | |
2232 | return pskb_may_pull(skb, skb_network_offset(skb) + len); | |
2233 | } | |
2234 | ||
1da177e4 LT |
2235 | /* |
2236 | * CPUs often take a performance hit when accessing unaligned memory | |
2237 | * locations. The actual performance hit varies, it can be small if the | |
2238 | * hardware handles it or large if we have to take an exception and fix it | |
2239 | * in software. | |
2240 | * | |
2241 | * Since an ethernet header is 14 bytes network drivers often end up with | |
2242 | * the IP header at an unaligned offset. The IP header can be aligned by | |
2243 | * shifting the start of the packet by 2 bytes. Drivers should do this | |
2244 | * with: | |
2245 | * | |
8660c124 | 2246 | * skb_reserve(skb, NET_IP_ALIGN); |
1da177e4 LT |
2247 | * |
2248 | * The downside to this alignment of the IP header is that the DMA is now | |
2249 | * unaligned. On some architectures the cost of an unaligned DMA is high | |
2250 | * and this cost outweighs the gains made by aligning the IP header. | |
8660c124 | 2251 | * |
1da177e4 LT |
2252 | * Since this trade off varies between architectures, we allow NET_IP_ALIGN |
2253 | * to be overridden. | |
2254 | */ | |
2255 | #ifndef NET_IP_ALIGN | |
2256 | #define NET_IP_ALIGN 2 | |
2257 | #endif | |
2258 | ||
025be81e AB |
2259 | /* |
2260 | * The networking layer reserves some headroom in skb data (via | |
2261 | * dev_alloc_skb). This is used to avoid having to reallocate skb data when | |
2262 | * the header has to grow. In the default case, if the header has to grow | |
d6301d3d | 2263 | * 32 bytes or less we avoid the reallocation. |
025be81e AB |
2264 | * |
2265 | * Unfortunately this headroom changes the DMA alignment of the resulting | |
2266 | * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive | |
2267 | * on some architectures. An architecture can override this value, | |
2268 | * perhaps setting it to a cacheline in size (since that will maintain | |
2269 | * cacheline alignment of the DMA). It must be a power of 2. | |
2270 | * | |
d6301d3d | 2271 | * Various parts of the networking layer expect at least 32 bytes of |
025be81e | 2272 | * headroom, you should not reduce this. |
5933dd2f ED |
2273 | * |
2274 | * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS) | |
2275 | * to reduce average number of cache lines per packet. | |
2276 | * get_rps_cpus() for example only access one 64 bytes aligned block : | |
18e8c134 | 2277 | * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8) |
025be81e AB |
2278 | */ |
2279 | #ifndef NET_SKB_PAD | |
5933dd2f | 2280 | #define NET_SKB_PAD max(32, L1_CACHE_BYTES) |
025be81e AB |
2281 | #endif |
2282 | ||
7965bd4d | 2283 | int ___pskb_trim(struct sk_buff *skb, unsigned int len); |
1da177e4 LT |
2284 | |
2285 | static inline void __skb_trim(struct sk_buff *skb, unsigned int len) | |
2286 | { | |
c4264f27 | 2287 | if (unlikely(skb_is_nonlinear(skb))) { |
3cc0e873 HX |
2288 | WARN_ON(1); |
2289 | return; | |
2290 | } | |
27a884dc ACM |
2291 | skb->len = len; |
2292 | skb_set_tail_pointer(skb, len); | |
1da177e4 LT |
2293 | } |
2294 | ||
7965bd4d | 2295 | void skb_trim(struct sk_buff *skb, unsigned int len); |
1da177e4 LT |
2296 | |
2297 | static inline int __pskb_trim(struct sk_buff *skb, unsigned int len) | |
2298 | { | |
3cc0e873 HX |
2299 | if (skb->data_len) |
2300 | return ___pskb_trim(skb, len); | |
2301 | __skb_trim(skb, len); | |
2302 | return 0; | |
1da177e4 LT |
2303 | } |
2304 | ||
2305 | static inline int pskb_trim(struct sk_buff *skb, unsigned int len) | |
2306 | { | |
2307 | return (len < skb->len) ? __pskb_trim(skb, len) : 0; | |
2308 | } | |
2309 | ||
e9fa4f7b HX |
2310 | /** |
2311 | * pskb_trim_unique - remove end from a paged unique (not cloned) buffer | |
2312 | * @skb: buffer to alter | |
2313 | * @len: new length | |
2314 | * | |
2315 | * This is identical to pskb_trim except that the caller knows that | |
2316 | * the skb is not cloned so we should never get an error due to out- | |
2317 | * of-memory. | |
2318 | */ | |
2319 | static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len) | |
2320 | { | |
2321 | int err = pskb_trim(skb, len); | |
2322 | BUG_ON(err); | |
2323 | } | |
2324 | ||
1da177e4 LT |
2325 | /** |
2326 | * skb_orphan - orphan a buffer | |
2327 | * @skb: buffer to orphan | |
2328 | * | |
2329 | * If a buffer currently has an owner then we call the owner's | |
2330 | * destructor function and make the @skb unowned. The buffer continues | |
2331 | * to exist but is no longer charged to its former owner. | |
2332 | */ | |
2333 | static inline void skb_orphan(struct sk_buff *skb) | |
2334 | { | |
c34a7612 | 2335 | if (skb->destructor) { |
1da177e4 | 2336 | skb->destructor(skb); |
c34a7612 ED |
2337 | skb->destructor = NULL; |
2338 | skb->sk = NULL; | |
376c7311 ED |
2339 | } else { |
2340 | BUG_ON(skb->sk); | |
c34a7612 | 2341 | } |
1da177e4 LT |
2342 | } |
2343 | ||
a353e0ce MT |
2344 | /** |
2345 | * skb_orphan_frags - orphan the frags contained in a buffer | |
2346 | * @skb: buffer to orphan frags from | |
2347 | * @gfp_mask: allocation mask for replacement pages | |
2348 | * | |
2349 | * For each frag in the SKB which needs a destructor (i.e. has an | |
2350 | * owner) create a copy of that frag and release the original | |
2351 | * page by calling the destructor. | |
2352 | */ | |
2353 | static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask) | |
2354 | { | |
2355 | if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY))) | |
2356 | return 0; | |
2357 | return skb_copy_ubufs(skb, gfp_mask); | |
2358 | } | |
2359 | ||
1da177e4 LT |
2360 | /** |
2361 | * __skb_queue_purge - empty a list | |
2362 | * @list: list to empty | |
2363 | * | |
2364 | * Delete all buffers on an &sk_buff list. Each buffer is removed from | |
2365 | * the list and one reference dropped. This function does not take the | |
2366 | * list lock and the caller must hold the relevant locks to use it. | |
2367 | */ | |
7965bd4d | 2368 | void skb_queue_purge(struct sk_buff_head *list); |
1da177e4 LT |
2369 | static inline void __skb_queue_purge(struct sk_buff_head *list) |
2370 | { | |
2371 | struct sk_buff *skb; | |
2372 | while ((skb = __skb_dequeue(list)) != NULL) | |
2373 | kfree_skb(skb); | |
2374 | } | |
2375 | ||
7965bd4d | 2376 | void *netdev_alloc_frag(unsigned int fragsz); |
1da177e4 | 2377 | |
7965bd4d JP |
2378 | struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length, |
2379 | gfp_t gfp_mask); | |
8af27456 CH |
2380 | |
2381 | /** | |
2382 | * netdev_alloc_skb - allocate an skbuff for rx on a specific device | |
2383 | * @dev: network device to receive on | |
2384 | * @length: length to allocate | |
2385 | * | |
2386 | * Allocate a new &sk_buff and assign it a usage count of one. The | |
2387 | * buffer has unspecified headroom built in. Users should allocate | |
2388 | * the headroom they think they need without accounting for the | |
2389 | * built in space. The built in space is used for optimisations. | |
2390 | * | |
2391 | * %NULL is returned if there is no free memory. Although this function | |
2392 | * allocates memory it can be called from an interrupt. | |
2393 | */ | |
2394 | static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev, | |
6f532612 | 2395 | unsigned int length) |
8af27456 CH |
2396 | { |
2397 | return __netdev_alloc_skb(dev, length, GFP_ATOMIC); | |
2398 | } | |
2399 | ||
6f532612 ED |
2400 | /* legacy helper around __netdev_alloc_skb() */ |
2401 | static inline struct sk_buff *__dev_alloc_skb(unsigned int length, | |
2402 | gfp_t gfp_mask) | |
2403 | { | |
2404 | return __netdev_alloc_skb(NULL, length, gfp_mask); | |
2405 | } | |
2406 | ||
2407 | /* legacy helper around netdev_alloc_skb() */ | |
2408 | static inline struct sk_buff *dev_alloc_skb(unsigned int length) | |
2409 | { | |
2410 | return netdev_alloc_skb(NULL, length); | |
2411 | } | |
2412 | ||
2413 | ||
4915a0de ED |
2414 | static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev, |
2415 | unsigned int length, gfp_t gfp) | |
61321bbd | 2416 | { |
4915a0de | 2417 | struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp); |
61321bbd ED |
2418 | |
2419 | if (NET_IP_ALIGN && skb) | |
2420 | skb_reserve(skb, NET_IP_ALIGN); | |
2421 | return skb; | |
2422 | } | |
2423 | ||
4915a0de ED |
2424 | static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev, |
2425 | unsigned int length) | |
2426 | { | |
2427 | return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC); | |
2428 | } | |
2429 | ||
181edb2b AD |
2430 | static inline void skb_free_frag(void *addr) |
2431 | { | |
2432 | __free_page_frag(addr); | |
2433 | } | |
2434 | ||
ffde7328 | 2435 | void *napi_alloc_frag(unsigned int fragsz); |
fd11a83d AD |
2436 | struct sk_buff *__napi_alloc_skb(struct napi_struct *napi, |
2437 | unsigned int length, gfp_t gfp_mask); | |
2438 | static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi, | |
2439 | unsigned int length) | |
2440 | { | |
2441 | return __napi_alloc_skb(napi, length, GFP_ATOMIC); | |
2442 | } | |
795bb1c0 JDB |
2443 | void napi_consume_skb(struct sk_buff *skb, int budget); |
2444 | ||
2445 | void __kfree_skb_flush(void); | |
15fad714 | 2446 | void __kfree_skb_defer(struct sk_buff *skb); |
ffde7328 | 2447 | |
71dfda58 AD |
2448 | /** |
2449 | * __dev_alloc_pages - allocate page for network Rx | |
2450 | * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx | |
2451 | * @order: size of the allocation | |
2452 | * | |
2453 | * Allocate a new page. | |
2454 | * | |
2455 | * %NULL is returned if there is no free memory. | |
2456 | */ | |
2457 | static inline struct page *__dev_alloc_pages(gfp_t gfp_mask, | |
2458 | unsigned int order) | |
2459 | { | |
2460 | /* This piece of code contains several assumptions. | |
2461 | * 1. This is for device Rx, therefor a cold page is preferred. | |
2462 | * 2. The expectation is the user wants a compound page. | |
2463 | * 3. If requesting a order 0 page it will not be compound | |
2464 | * due to the check to see if order has a value in prep_new_page | |
2465 | * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to | |
2466 | * code in gfp_to_alloc_flags that should be enforcing this. | |
2467 | */ | |
2468 | gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC; | |
2469 | ||
2470 | return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order); | |
2471 | } | |
2472 | ||
2473 | static inline struct page *dev_alloc_pages(unsigned int order) | |
2474 | { | |
95829b3a | 2475 | return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order); |
71dfda58 AD |
2476 | } |
2477 | ||
2478 | /** | |
2479 | * __dev_alloc_page - allocate a page for network Rx | |
2480 | * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx | |
2481 | * | |
2482 | * Allocate a new page. | |
2483 | * | |
2484 | * %NULL is returned if there is no free memory. | |
2485 | */ | |
2486 | static inline struct page *__dev_alloc_page(gfp_t gfp_mask) | |
2487 | { | |
2488 | return __dev_alloc_pages(gfp_mask, 0); | |
2489 | } | |
2490 | ||
2491 | static inline struct page *dev_alloc_page(void) | |
2492 | { | |
95829b3a | 2493 | return dev_alloc_pages(0); |
71dfda58 AD |
2494 | } |
2495 | ||
0614002b MG |
2496 | /** |
2497 | * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page | |
2498 | * @page: The page that was allocated from skb_alloc_page | |
2499 | * @skb: The skb that may need pfmemalloc set | |
2500 | */ | |
2501 | static inline void skb_propagate_pfmemalloc(struct page *page, | |
2502 | struct sk_buff *skb) | |
2503 | { | |
2f064f34 | 2504 | if (page_is_pfmemalloc(page)) |
0614002b MG |
2505 | skb->pfmemalloc = true; |
2506 | } | |
2507 | ||
131ea667 | 2508 | /** |
e227867f | 2509 | * skb_frag_page - retrieve the page referred to by a paged fragment |
131ea667 IC |
2510 | * @frag: the paged fragment |
2511 | * | |
2512 | * Returns the &struct page associated with @frag. | |
2513 | */ | |
2514 | static inline struct page *skb_frag_page(const skb_frag_t *frag) | |
2515 | { | |
a8605c60 | 2516 | return frag->page.p; |
131ea667 IC |
2517 | } |
2518 | ||
2519 | /** | |
2520 | * __skb_frag_ref - take an addition reference on a paged fragment. | |
2521 | * @frag: the paged fragment | |
2522 | * | |
2523 | * Takes an additional reference on the paged fragment @frag. | |
2524 | */ | |
2525 | static inline void __skb_frag_ref(skb_frag_t *frag) | |
2526 | { | |
2527 | get_page(skb_frag_page(frag)); | |
2528 | } | |
2529 | ||
2530 | /** | |
2531 | * skb_frag_ref - take an addition reference on a paged fragment of an skb. | |
2532 | * @skb: the buffer | |
2533 | * @f: the fragment offset. | |
2534 | * | |
2535 | * Takes an additional reference on the @f'th paged fragment of @skb. | |
2536 | */ | |
2537 | static inline void skb_frag_ref(struct sk_buff *skb, int f) | |
2538 | { | |
2539 | __skb_frag_ref(&skb_shinfo(skb)->frags[f]); | |
2540 | } | |
2541 | ||
2542 | /** | |
2543 | * __skb_frag_unref - release a reference on a paged fragment. | |
2544 | * @frag: the paged fragment | |
2545 | * | |
2546 | * Releases a reference on the paged fragment @frag. | |
2547 | */ | |
2548 | static inline void __skb_frag_unref(skb_frag_t *frag) | |
2549 | { | |
2550 | put_page(skb_frag_page(frag)); | |
2551 | } | |
2552 | ||
2553 | /** | |
2554 | * skb_frag_unref - release a reference on a paged fragment of an skb. | |
2555 | * @skb: the buffer | |
2556 | * @f: the fragment offset | |
2557 | * | |
2558 | * Releases a reference on the @f'th paged fragment of @skb. | |
2559 | */ | |
2560 | static inline void skb_frag_unref(struct sk_buff *skb, int f) | |
2561 | { | |
2562 | __skb_frag_unref(&skb_shinfo(skb)->frags[f]); | |
2563 | } | |
2564 | ||
2565 | /** | |
2566 | * skb_frag_address - gets the address of the data contained in a paged fragment | |
2567 | * @frag: the paged fragment buffer | |
2568 | * | |
2569 | * Returns the address of the data within @frag. The page must already | |
2570 | * be mapped. | |
2571 | */ | |
2572 | static inline void *skb_frag_address(const skb_frag_t *frag) | |
2573 | { | |
2574 | return page_address(skb_frag_page(frag)) + frag->page_offset; | |
2575 | } | |
2576 | ||
2577 | /** | |
2578 | * skb_frag_address_safe - gets the address of the data contained in a paged fragment | |
2579 | * @frag: the paged fragment buffer | |
2580 | * | |
2581 | * Returns the address of the data within @frag. Checks that the page | |
2582 | * is mapped and returns %NULL otherwise. | |
2583 | */ | |
2584 | static inline void *skb_frag_address_safe(const skb_frag_t *frag) | |
2585 | { | |
2586 | void *ptr = page_address(skb_frag_page(frag)); | |
2587 | if (unlikely(!ptr)) | |
2588 | return NULL; | |
2589 | ||
2590 | return ptr + frag->page_offset; | |
2591 | } | |
2592 | ||
2593 | /** | |
2594 | * __skb_frag_set_page - sets the page contained in a paged fragment | |
2595 | * @frag: the paged fragment | |
2596 | * @page: the page to set | |
2597 | * | |
2598 | * Sets the fragment @frag to contain @page. | |
2599 | */ | |
2600 | static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page) | |
2601 | { | |
a8605c60 | 2602 | frag->page.p = page; |
131ea667 IC |
2603 | } |
2604 | ||
2605 | /** | |
2606 | * skb_frag_set_page - sets the page contained in a paged fragment of an skb | |
2607 | * @skb: the buffer | |
2608 | * @f: the fragment offset | |
2609 | * @page: the page to set | |
2610 | * | |
2611 | * Sets the @f'th fragment of @skb to contain @page. | |
2612 | */ | |
2613 | static inline void skb_frag_set_page(struct sk_buff *skb, int f, | |
2614 | struct page *page) | |
2615 | { | |
2616 | __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page); | |
2617 | } | |
2618 | ||
400dfd3a ED |
2619 | bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio); |
2620 | ||
131ea667 IC |
2621 | /** |
2622 | * skb_frag_dma_map - maps a paged fragment via the DMA API | |
f83347df | 2623 | * @dev: the device to map the fragment to |
131ea667 IC |
2624 | * @frag: the paged fragment to map |
2625 | * @offset: the offset within the fragment (starting at the | |
2626 | * fragment's own offset) | |
2627 | * @size: the number of bytes to map | |
f83347df | 2628 | * @dir: the direction of the mapping (%PCI_DMA_*) |
131ea667 IC |
2629 | * |
2630 | * Maps the page associated with @frag to @device. | |
2631 | */ | |
2632 | static inline dma_addr_t skb_frag_dma_map(struct device *dev, | |
2633 | const skb_frag_t *frag, | |
2634 | size_t offset, size_t size, | |
2635 | enum dma_data_direction dir) | |
2636 | { | |
2637 | return dma_map_page(dev, skb_frag_page(frag), | |
2638 | frag->page_offset + offset, size, dir); | |
2639 | } | |
2640 | ||
117632e6 ED |
2641 | static inline struct sk_buff *pskb_copy(struct sk_buff *skb, |
2642 | gfp_t gfp_mask) | |
2643 | { | |
2644 | return __pskb_copy(skb, skb_headroom(skb), gfp_mask); | |
2645 | } | |
2646 | ||
bad93e9d OP |
2647 | |
2648 | static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb, | |
2649 | gfp_t gfp_mask) | |
2650 | { | |
2651 | return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true); | |
2652 | } | |
2653 | ||
2654 | ||
334a8132 PM |
2655 | /** |
2656 | * skb_clone_writable - is the header of a clone writable | |
2657 | * @skb: buffer to check | |
2658 | * @len: length up to which to write | |
2659 | * | |
2660 | * Returns true if modifying the header part of the cloned buffer | |
2661 | * does not requires the data to be copied. | |
2662 | */ | |
05bdd2f1 | 2663 | static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len) |
334a8132 PM |
2664 | { |
2665 | return !skb_header_cloned(skb) && | |
2666 | skb_headroom(skb) + len <= skb->hdr_len; | |
2667 | } | |
2668 | ||
3697649f DB |
2669 | static inline int skb_try_make_writable(struct sk_buff *skb, |
2670 | unsigned int write_len) | |
2671 | { | |
2672 | return skb_cloned(skb) && !skb_clone_writable(skb, write_len) && | |
2673 | pskb_expand_head(skb, 0, 0, GFP_ATOMIC); | |
2674 | } | |
2675 | ||
d9cc2048 HX |
2676 | static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom, |
2677 | int cloned) | |
2678 | { | |
2679 | int delta = 0; | |
2680 | ||
d9cc2048 HX |
2681 | if (headroom > skb_headroom(skb)) |
2682 | delta = headroom - skb_headroom(skb); | |
2683 | ||
2684 | if (delta || cloned) | |
2685 | return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0, | |
2686 | GFP_ATOMIC); | |
2687 | return 0; | |
2688 | } | |
2689 | ||
1da177e4 LT |
2690 | /** |
2691 | * skb_cow - copy header of skb when it is required | |
2692 | * @skb: buffer to cow | |
2693 | * @headroom: needed headroom | |
2694 | * | |
2695 | * If the skb passed lacks sufficient headroom or its data part | |
2696 | * is shared, data is reallocated. If reallocation fails, an error | |
2697 | * is returned and original skb is not changed. | |
2698 | * | |
2699 | * The result is skb with writable area skb->head...skb->tail | |
2700 | * and at least @headroom of space at head. | |
2701 | */ | |
2702 | static inline int skb_cow(struct sk_buff *skb, unsigned int headroom) | |
2703 | { | |
d9cc2048 HX |
2704 | return __skb_cow(skb, headroom, skb_cloned(skb)); |
2705 | } | |
1da177e4 | 2706 | |
d9cc2048 HX |
2707 | /** |
2708 | * skb_cow_head - skb_cow but only making the head writable | |
2709 | * @skb: buffer to cow | |
2710 | * @headroom: needed headroom | |
2711 | * | |
2712 | * This function is identical to skb_cow except that we replace the | |
2713 | * skb_cloned check by skb_header_cloned. It should be used when | |
2714 | * you only need to push on some header and do not need to modify | |
2715 | * the data. | |
2716 | */ | |
2717 | static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom) | |
2718 | { | |
2719 | return __skb_cow(skb, headroom, skb_header_cloned(skb)); | |
1da177e4 LT |
2720 | } |
2721 | ||
2722 | /** | |
2723 | * skb_padto - pad an skbuff up to a minimal size | |
2724 | * @skb: buffer to pad | |
2725 | * @len: minimal length | |
2726 | * | |
2727 | * Pads up a buffer to ensure the trailing bytes exist and are | |
2728 | * blanked. If the buffer already contains sufficient data it | |
5b057c6b HX |
2729 | * is untouched. Otherwise it is extended. Returns zero on |
2730 | * success. The skb is freed on error. | |
1da177e4 | 2731 | */ |
5b057c6b | 2732 | static inline int skb_padto(struct sk_buff *skb, unsigned int len) |
1da177e4 LT |
2733 | { |
2734 | unsigned int size = skb->len; | |
2735 | if (likely(size >= len)) | |
5b057c6b | 2736 | return 0; |
987c402a | 2737 | return skb_pad(skb, len - size); |
1da177e4 LT |
2738 | } |
2739 | ||
9c0c1124 AD |
2740 | /** |
2741 | * skb_put_padto - increase size and pad an skbuff up to a minimal size | |
2742 | * @skb: buffer to pad | |
2743 | * @len: minimal length | |
2744 | * | |
2745 | * Pads up a buffer to ensure the trailing bytes exist and are | |
2746 | * blanked. If the buffer already contains sufficient data it | |
2747 | * is untouched. Otherwise it is extended. Returns zero on | |
2748 | * success. The skb is freed on error. | |
2749 | */ | |
2750 | static inline int skb_put_padto(struct sk_buff *skb, unsigned int len) | |
2751 | { | |
2752 | unsigned int size = skb->len; | |
2753 | ||
2754 | if (unlikely(size < len)) { | |
2755 | len -= size; | |
2756 | if (skb_pad(skb, len)) | |
2757 | return -ENOMEM; | |
2758 | __skb_put(skb, len); | |
2759 | } | |
2760 | return 0; | |
2761 | } | |
2762 | ||
1da177e4 | 2763 | static inline int skb_add_data(struct sk_buff *skb, |
af2b040e | 2764 | struct iov_iter *from, int copy) |
1da177e4 LT |
2765 | { |
2766 | const int off = skb->len; | |
2767 | ||
2768 | if (skb->ip_summed == CHECKSUM_NONE) { | |
af2b040e AV |
2769 | __wsum csum = 0; |
2770 | if (csum_and_copy_from_iter(skb_put(skb, copy), copy, | |
2771 | &csum, from) == copy) { | |
1da177e4 LT |
2772 | skb->csum = csum_block_add(skb->csum, csum, off); |
2773 | return 0; | |
2774 | } | |
af2b040e | 2775 | } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy) |
1da177e4 LT |
2776 | return 0; |
2777 | ||
2778 | __skb_trim(skb, off); | |
2779 | return -EFAULT; | |
2780 | } | |
2781 | ||
38ba0a65 ED |
2782 | static inline bool skb_can_coalesce(struct sk_buff *skb, int i, |
2783 | const struct page *page, int off) | |
1da177e4 LT |
2784 | { |
2785 | if (i) { | |
9e903e08 | 2786 | const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1]; |
1da177e4 | 2787 | |
ea2ab693 | 2788 | return page == skb_frag_page(frag) && |
9e903e08 | 2789 | off == frag->page_offset + skb_frag_size(frag); |
1da177e4 | 2790 | } |
38ba0a65 | 2791 | return false; |
1da177e4 LT |
2792 | } |
2793 | ||
364c6bad HX |
2794 | static inline int __skb_linearize(struct sk_buff *skb) |
2795 | { | |
2796 | return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM; | |
2797 | } | |
2798 | ||
1da177e4 LT |
2799 | /** |
2800 | * skb_linearize - convert paged skb to linear one | |
2801 | * @skb: buffer to linarize | |
1da177e4 LT |
2802 | * |
2803 | * If there is no free memory -ENOMEM is returned, otherwise zero | |
2804 | * is returned and the old skb data released. | |
2805 | */ | |
364c6bad HX |
2806 | static inline int skb_linearize(struct sk_buff *skb) |
2807 | { | |
2808 | return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0; | |
2809 | } | |
2810 | ||
cef401de ED |
2811 | /** |
2812 | * skb_has_shared_frag - can any frag be overwritten | |
2813 | * @skb: buffer to test | |
2814 | * | |
2815 | * Return true if the skb has at least one frag that might be modified | |
2816 | * by an external entity (as in vmsplice()/sendfile()) | |
2817 | */ | |
2818 | static inline bool skb_has_shared_frag(const struct sk_buff *skb) | |
2819 | { | |
c9af6db4 PS |
2820 | return skb_is_nonlinear(skb) && |
2821 | skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG; | |
cef401de ED |
2822 | } |
2823 | ||
364c6bad HX |
2824 | /** |
2825 | * skb_linearize_cow - make sure skb is linear and writable | |
2826 | * @skb: buffer to process | |
2827 | * | |
2828 | * If there is no free memory -ENOMEM is returned, otherwise zero | |
2829 | * is returned and the old skb data released. | |
2830 | */ | |
2831 | static inline int skb_linearize_cow(struct sk_buff *skb) | |
1da177e4 | 2832 | { |
364c6bad HX |
2833 | return skb_is_nonlinear(skb) || skb_cloned(skb) ? |
2834 | __skb_linearize(skb) : 0; | |
1da177e4 LT |
2835 | } |
2836 | ||
2837 | /** | |
2838 | * skb_postpull_rcsum - update checksum for received skb after pull | |
2839 | * @skb: buffer to update | |
2840 | * @start: start of data before pull | |
2841 | * @len: length of data pulled | |
2842 | * | |
2843 | * After doing a pull on a received packet, you need to call this to | |
84fa7933 PM |
2844 | * update the CHECKSUM_COMPLETE checksum, or set ip_summed to |
2845 | * CHECKSUM_NONE so that it can be recomputed from scratch. | |
1da177e4 LT |
2846 | */ |
2847 | ||
2848 | static inline void skb_postpull_rcsum(struct sk_buff *skb, | |
cbb042f9 | 2849 | const void *start, unsigned int len) |
1da177e4 | 2850 | { |
84fa7933 | 2851 | if (skb->ip_summed == CHECKSUM_COMPLETE) |
1da177e4 | 2852 | skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0)); |
6ae459bd | 2853 | else if (skb->ip_summed == CHECKSUM_PARTIAL && |
31b33dfb | 2854 | skb_checksum_start_offset(skb) < 0) |
6ae459bd | 2855 | skb->ip_summed = CHECKSUM_NONE; |
1da177e4 LT |
2856 | } |
2857 | ||
cbb042f9 HX |
2858 | unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len); |
2859 | ||
f8ffad69 DB |
2860 | static inline void skb_postpush_rcsum(struct sk_buff *skb, |
2861 | const void *start, unsigned int len) | |
2862 | { | |
2863 | /* For performing the reverse operation to skb_postpull_rcsum(), | |
2864 | * we can instead of ... | |
2865 | * | |
2866 | * skb->csum = csum_add(skb->csum, csum_partial(start, len, 0)); | |
2867 | * | |
2868 | * ... just use this equivalent version here to save a few | |
2869 | * instructions. Feeding csum of 0 in csum_partial() and later | |
2870 | * on adding skb->csum is equivalent to feed skb->csum in the | |
2871 | * first place. | |
2872 | */ | |
2873 | if (skb->ip_summed == CHECKSUM_COMPLETE) | |
2874 | skb->csum = csum_partial(start, len, skb->csum); | |
2875 | } | |
2876 | ||
7ce5a27f DM |
2877 | /** |
2878 | * pskb_trim_rcsum - trim received skb and update checksum | |
2879 | * @skb: buffer to trim | |
2880 | * @len: new length | |
2881 | * | |
2882 | * This is exactly the same as pskb_trim except that it ensures the | |
2883 | * checksum of received packets are still valid after the operation. | |
2884 | */ | |
2885 | ||
2886 | static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len) | |
2887 | { | |
2888 | if (likely(len >= skb->len)) | |
2889 | return 0; | |
2890 | if (skb->ip_summed == CHECKSUM_COMPLETE) | |
2891 | skb->ip_summed = CHECKSUM_NONE; | |
2892 | return __pskb_trim(skb, len); | |
2893 | } | |
2894 | ||
1da177e4 LT |
2895 | #define skb_queue_walk(queue, skb) \ |
2896 | for (skb = (queue)->next; \ | |
a1e4891f | 2897 | skb != (struct sk_buff *)(queue); \ |
1da177e4 LT |
2898 | skb = skb->next) |
2899 | ||
46f8914e JC |
2900 | #define skb_queue_walk_safe(queue, skb, tmp) \ |
2901 | for (skb = (queue)->next, tmp = skb->next; \ | |
2902 | skb != (struct sk_buff *)(queue); \ | |
2903 | skb = tmp, tmp = skb->next) | |
2904 | ||
1164f52a | 2905 | #define skb_queue_walk_from(queue, skb) \ |
a1e4891f | 2906 | for (; skb != (struct sk_buff *)(queue); \ |
1164f52a DM |
2907 | skb = skb->next) |
2908 | ||
2909 | #define skb_queue_walk_from_safe(queue, skb, tmp) \ | |
2910 | for (tmp = skb->next; \ | |
2911 | skb != (struct sk_buff *)(queue); \ | |
2912 | skb = tmp, tmp = skb->next) | |
2913 | ||
300ce174 SH |
2914 | #define skb_queue_reverse_walk(queue, skb) \ |
2915 | for (skb = (queue)->prev; \ | |
a1e4891f | 2916 | skb != (struct sk_buff *)(queue); \ |
300ce174 SH |
2917 | skb = skb->prev) |
2918 | ||
686a2955 DM |
2919 | #define skb_queue_reverse_walk_safe(queue, skb, tmp) \ |
2920 | for (skb = (queue)->prev, tmp = skb->prev; \ | |
2921 | skb != (struct sk_buff *)(queue); \ | |
2922 | skb = tmp, tmp = skb->prev) | |
2923 | ||
2924 | #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \ | |
2925 | for (tmp = skb->prev; \ | |
2926 | skb != (struct sk_buff *)(queue); \ | |
2927 | skb = tmp, tmp = skb->prev) | |
1da177e4 | 2928 | |
21dc3301 | 2929 | static inline bool skb_has_frag_list(const struct sk_buff *skb) |
ee039871 DM |
2930 | { |
2931 | return skb_shinfo(skb)->frag_list != NULL; | |
2932 | } | |
2933 | ||
2934 | static inline void skb_frag_list_init(struct sk_buff *skb) | |
2935 | { | |
2936 | skb_shinfo(skb)->frag_list = NULL; | |
2937 | } | |
2938 | ||
ee039871 DM |
2939 | #define skb_walk_frags(skb, iter) \ |
2940 | for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next) | |
2941 | ||
ea3793ee RW |
2942 | |
2943 | int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p, | |
2944 | const struct sk_buff *skb); | |
2945 | struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags, | |
2946 | int *peeked, int *off, int *err, | |
2947 | struct sk_buff **last); | |
7965bd4d JP |
2948 | struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags, |
2949 | int *peeked, int *off, int *err); | |
2950 | struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock, | |
2951 | int *err); | |
2952 | unsigned int datagram_poll(struct file *file, struct socket *sock, | |
2953 | struct poll_table_struct *wait); | |
c0371da6 AV |
2954 | int skb_copy_datagram_iter(const struct sk_buff *from, int offset, |
2955 | struct iov_iter *to, int size); | |
51f3d02b DM |
2956 | static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset, |
2957 | struct msghdr *msg, int size) | |
2958 | { | |
e5a4b0bb | 2959 | return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size); |
227158db | 2960 | } |
e5a4b0bb AV |
2961 | int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen, |
2962 | struct msghdr *msg); | |
3a654f97 AV |
2963 | int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset, |
2964 | struct iov_iter *from, int len); | |
3a654f97 | 2965 | int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm); |
7965bd4d | 2966 | void skb_free_datagram(struct sock *sk, struct sk_buff *skb); |
627d2d6b | 2967 | void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len); |
2968 | static inline void skb_free_datagram_locked(struct sock *sk, | |
2969 | struct sk_buff *skb) | |
2970 | { | |
2971 | __skb_free_datagram_locked(sk, skb, 0); | |
2972 | } | |
7965bd4d | 2973 | int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags); |
7965bd4d JP |
2974 | int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len); |
2975 | int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len); | |
2976 | __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to, | |
2977 | int len, __wsum csum); | |
a60e3cc7 HFS |
2978 | ssize_t skb_socket_splice(struct sock *sk, |
2979 | struct pipe_inode_info *pipe, | |
2980 | struct splice_pipe_desc *spd); | |
2981 | int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset, | |
7965bd4d | 2982 | struct pipe_inode_info *pipe, unsigned int len, |
a60e3cc7 HFS |
2983 | unsigned int flags, |
2984 | ssize_t (*splice_cb)(struct sock *, | |
2985 | struct pipe_inode_info *, | |
2986 | struct splice_pipe_desc *)); | |
7965bd4d | 2987 | void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to); |
af2806f8 | 2988 | unsigned int skb_zerocopy_headlen(const struct sk_buff *from); |
36d5fe6a ZK |
2989 | int skb_zerocopy(struct sk_buff *to, struct sk_buff *from, |
2990 | int len, int hlen); | |
7965bd4d JP |
2991 | void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len); |
2992 | int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen); | |
2993 | void skb_scrub_packet(struct sk_buff *skb, bool xnet); | |
de960aa9 | 2994 | unsigned int skb_gso_transport_seglen(const struct sk_buff *skb); |
ae7ef81e | 2995 | bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu); |
7965bd4d | 2996 | struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features); |
0d5501c1 | 2997 | struct sk_buff *skb_vlan_untag(struct sk_buff *skb); |
e2195121 | 2998 | int skb_ensure_writable(struct sk_buff *skb, int write_len); |
93515d53 JP |
2999 | int skb_vlan_pop(struct sk_buff *skb); |
3000 | int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci); | |
6fa01ccd SV |
3001 | struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy, |
3002 | gfp_t gfp); | |
20380731 | 3003 | |
6ce8e9ce AV |
3004 | static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len) |
3005 | { | |
21226abb | 3006 | return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT; |
6ce8e9ce AV |
3007 | } |
3008 | ||
7eab8d9e AV |
3009 | static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len) |
3010 | { | |
e5a4b0bb | 3011 | return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT; |
7eab8d9e AV |
3012 | } |
3013 | ||
2817a336 DB |
3014 | struct skb_checksum_ops { |
3015 | __wsum (*update)(const void *mem, int len, __wsum wsum); | |
3016 | __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len); | |
3017 | }; | |
3018 | ||
3019 | __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len, | |
3020 | __wsum csum, const struct skb_checksum_ops *ops); | |
3021 | __wsum skb_checksum(const struct sk_buff *skb, int offset, int len, | |
3022 | __wsum csum); | |
3023 | ||
1e98a0f0 ED |
3024 | static inline void * __must_check |
3025 | __skb_header_pointer(const struct sk_buff *skb, int offset, | |
3026 | int len, void *data, int hlen, void *buffer) | |
1da177e4 | 3027 | { |
55820ee2 | 3028 | if (hlen - offset >= len) |
690e36e7 | 3029 | return data + offset; |
1da177e4 | 3030 | |
690e36e7 DM |
3031 | if (!skb || |
3032 | skb_copy_bits(skb, offset, buffer, len) < 0) | |
1da177e4 LT |
3033 | return NULL; |
3034 | ||
3035 | return buffer; | |
3036 | } | |
3037 | ||
1e98a0f0 ED |
3038 | static inline void * __must_check |
3039 | skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer) | |
690e36e7 DM |
3040 | { |
3041 | return __skb_header_pointer(skb, offset, len, skb->data, | |
3042 | skb_headlen(skb), buffer); | |
3043 | } | |
3044 | ||
4262e5cc DB |
3045 | /** |
3046 | * skb_needs_linearize - check if we need to linearize a given skb | |
3047 | * depending on the given device features. | |
3048 | * @skb: socket buffer to check | |
3049 | * @features: net device features | |
3050 | * | |
3051 | * Returns true if either: | |
3052 | * 1. skb has frag_list and the device doesn't support FRAGLIST, or | |
3053 | * 2. skb is fragmented and the device does not support SG. | |
3054 | */ | |
3055 | static inline bool skb_needs_linearize(struct sk_buff *skb, | |
3056 | netdev_features_t features) | |
3057 | { | |
3058 | return skb_is_nonlinear(skb) && | |
3059 | ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) || | |
3060 | (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG))); | |
3061 | } | |
3062 | ||
d626f62b ACM |
3063 | static inline void skb_copy_from_linear_data(const struct sk_buff *skb, |
3064 | void *to, | |
3065 | const unsigned int len) | |
3066 | { | |
3067 | memcpy(to, skb->data, len); | |
3068 | } | |
3069 | ||
3070 | static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb, | |
3071 | const int offset, void *to, | |
3072 | const unsigned int len) | |
3073 | { | |
3074 | memcpy(to, skb->data + offset, len); | |
3075 | } | |
3076 | ||
27d7ff46 ACM |
3077 | static inline void skb_copy_to_linear_data(struct sk_buff *skb, |
3078 | const void *from, | |
3079 | const unsigned int len) | |
3080 | { | |
3081 | memcpy(skb->data, from, len); | |
3082 | } | |
3083 | ||
3084 | static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb, | |
3085 | const int offset, | |
3086 | const void *from, | |
3087 | const unsigned int len) | |
3088 | { | |
3089 | memcpy(skb->data + offset, from, len); | |
3090 | } | |
3091 | ||
7965bd4d | 3092 | void skb_init(void); |
1da177e4 | 3093 | |
ac45f602 PO |
3094 | static inline ktime_t skb_get_ktime(const struct sk_buff *skb) |
3095 | { | |
3096 | return skb->tstamp; | |
3097 | } | |
3098 | ||
a61bbcf2 PM |
3099 | /** |
3100 | * skb_get_timestamp - get timestamp from a skb | |
3101 | * @skb: skb to get stamp from | |
3102 | * @stamp: pointer to struct timeval to store stamp in | |
3103 | * | |
3104 | * Timestamps are stored in the skb as offsets to a base timestamp. | |
3105 | * This function converts the offset back to a struct timeval and stores | |
3106 | * it in stamp. | |
3107 | */ | |
ac45f602 PO |
3108 | static inline void skb_get_timestamp(const struct sk_buff *skb, |
3109 | struct timeval *stamp) | |
a61bbcf2 | 3110 | { |
b7aa0bf7 | 3111 | *stamp = ktime_to_timeval(skb->tstamp); |
a61bbcf2 PM |
3112 | } |
3113 | ||
ac45f602 PO |
3114 | static inline void skb_get_timestampns(const struct sk_buff *skb, |
3115 | struct timespec *stamp) | |
3116 | { | |
3117 | *stamp = ktime_to_timespec(skb->tstamp); | |
3118 | } | |
3119 | ||
b7aa0bf7 | 3120 | static inline void __net_timestamp(struct sk_buff *skb) |
a61bbcf2 | 3121 | { |
b7aa0bf7 | 3122 | skb->tstamp = ktime_get_real(); |
a61bbcf2 PM |
3123 | } |
3124 | ||
164891aa SH |
3125 | static inline ktime_t net_timedelta(ktime_t t) |
3126 | { | |
3127 | return ktime_sub(ktime_get_real(), t); | |
3128 | } | |
3129 | ||
b9ce204f IJ |
3130 | static inline ktime_t net_invalid_timestamp(void) |
3131 | { | |
3132 | return ktime_set(0, 0); | |
3133 | } | |
a61bbcf2 | 3134 | |
62bccb8c AD |
3135 | struct sk_buff *skb_clone_sk(struct sk_buff *skb); |
3136 | ||
c1f19b51 RC |
3137 | #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING |
3138 | ||
7965bd4d JP |
3139 | void skb_clone_tx_timestamp(struct sk_buff *skb); |
3140 | bool skb_defer_rx_timestamp(struct sk_buff *skb); | |
c1f19b51 RC |
3141 | |
3142 | #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */ | |
3143 | ||
3144 | static inline void skb_clone_tx_timestamp(struct sk_buff *skb) | |
3145 | { | |
3146 | } | |
3147 | ||
3148 | static inline bool skb_defer_rx_timestamp(struct sk_buff *skb) | |
3149 | { | |
3150 | return false; | |
3151 | } | |
3152 | ||
3153 | #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */ | |
3154 | ||
3155 | /** | |
3156 | * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps | |
3157 | * | |
da92b194 RC |
3158 | * PHY drivers may accept clones of transmitted packets for |
3159 | * timestamping via their phy_driver.txtstamp method. These drivers | |
7a76a021 BP |
3160 | * must call this function to return the skb back to the stack with a |
3161 | * timestamp. | |
da92b194 | 3162 | * |
c1f19b51 | 3163 | * @skb: clone of the the original outgoing packet |
7a76a021 | 3164 | * @hwtstamps: hardware time stamps |
c1f19b51 RC |
3165 | * |
3166 | */ | |
3167 | void skb_complete_tx_timestamp(struct sk_buff *skb, | |
3168 | struct skb_shared_hwtstamps *hwtstamps); | |
3169 | ||
e7fd2885 WB |
3170 | void __skb_tstamp_tx(struct sk_buff *orig_skb, |
3171 | struct skb_shared_hwtstamps *hwtstamps, | |
3172 | struct sock *sk, int tstype); | |
3173 | ||
ac45f602 PO |
3174 | /** |
3175 | * skb_tstamp_tx - queue clone of skb with send time stamps | |
3176 | * @orig_skb: the original outgoing packet | |
3177 | * @hwtstamps: hardware time stamps, may be NULL if not available | |
3178 | * | |
3179 | * If the skb has a socket associated, then this function clones the | |
3180 | * skb (thus sharing the actual data and optional structures), stores | |
3181 | * the optional hardware time stamping information (if non NULL) or | |
3182 | * generates a software time stamp (otherwise), then queues the clone | |
3183 | * to the error queue of the socket. Errors are silently ignored. | |
3184 | */ | |
7965bd4d JP |
3185 | void skb_tstamp_tx(struct sk_buff *orig_skb, |
3186 | struct skb_shared_hwtstamps *hwtstamps); | |
ac45f602 | 3187 | |
4507a715 RC |
3188 | static inline void sw_tx_timestamp(struct sk_buff *skb) |
3189 | { | |
2244d07b OH |
3190 | if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP && |
3191 | !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS)) | |
4507a715 RC |
3192 | skb_tstamp_tx(skb, NULL); |
3193 | } | |
3194 | ||
3195 | /** | |
3196 | * skb_tx_timestamp() - Driver hook for transmit timestamping | |
3197 | * | |
3198 | * Ethernet MAC Drivers should call this function in their hard_xmit() | |
4ff75b7c | 3199 | * function immediately before giving the sk_buff to the MAC hardware. |
4507a715 | 3200 | * |
73409f3b DM |
3201 | * Specifically, one should make absolutely sure that this function is |
3202 | * called before TX completion of this packet can trigger. Otherwise | |
3203 | * the packet could potentially already be freed. | |
3204 | * | |
4507a715 RC |
3205 | * @skb: A socket buffer. |
3206 | */ | |
3207 | static inline void skb_tx_timestamp(struct sk_buff *skb) | |
3208 | { | |
c1f19b51 | 3209 | skb_clone_tx_timestamp(skb); |
4507a715 RC |
3210 | sw_tx_timestamp(skb); |
3211 | } | |
3212 | ||
6e3e939f JB |
3213 | /** |
3214 | * skb_complete_wifi_ack - deliver skb with wifi status | |
3215 | * | |
3216 | * @skb: the original outgoing packet | |
3217 | * @acked: ack status | |
3218 | * | |
3219 | */ | |
3220 | void skb_complete_wifi_ack(struct sk_buff *skb, bool acked); | |
3221 | ||
7965bd4d JP |
3222 | __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len); |
3223 | __sum16 __skb_checksum_complete(struct sk_buff *skb); | |
fb286bb2 | 3224 | |
60476372 HX |
3225 | static inline int skb_csum_unnecessary(const struct sk_buff *skb) |
3226 | { | |
6edec0e6 TH |
3227 | return ((skb->ip_summed == CHECKSUM_UNNECESSARY) || |
3228 | skb->csum_valid || | |
3229 | (skb->ip_summed == CHECKSUM_PARTIAL && | |
3230 | skb_checksum_start_offset(skb) >= 0)); | |
60476372 HX |
3231 | } |
3232 | ||
fb286bb2 HX |
3233 | /** |
3234 | * skb_checksum_complete - Calculate checksum of an entire packet | |
3235 | * @skb: packet to process | |
3236 | * | |
3237 | * This function calculates the checksum over the entire packet plus | |
3238 | * the value of skb->csum. The latter can be used to supply the | |
3239 | * checksum of a pseudo header as used by TCP/UDP. It returns the | |
3240 | * checksum. | |
3241 | * | |
3242 | * For protocols that contain complete checksums such as ICMP/TCP/UDP, | |
3243 | * this function can be used to verify that checksum on received | |
3244 | * packets. In that case the function should return zero if the | |
3245 | * checksum is correct. In particular, this function will return zero | |
3246 | * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the | |
3247 | * hardware has already verified the correctness of the checksum. | |
3248 | */ | |
4381ca3c | 3249 | static inline __sum16 skb_checksum_complete(struct sk_buff *skb) |
fb286bb2 | 3250 | { |
60476372 HX |
3251 | return skb_csum_unnecessary(skb) ? |
3252 | 0 : __skb_checksum_complete(skb); | |
fb286bb2 HX |
3253 | } |
3254 | ||
77cffe23 TH |
3255 | static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb) |
3256 | { | |
3257 | if (skb->ip_summed == CHECKSUM_UNNECESSARY) { | |
3258 | if (skb->csum_level == 0) | |
3259 | skb->ip_summed = CHECKSUM_NONE; | |
3260 | else | |
3261 | skb->csum_level--; | |
3262 | } | |
3263 | } | |
3264 | ||
3265 | static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb) | |
3266 | { | |
3267 | if (skb->ip_summed == CHECKSUM_UNNECESSARY) { | |
3268 | if (skb->csum_level < SKB_MAX_CSUM_LEVEL) | |
3269 | skb->csum_level++; | |
3270 | } else if (skb->ip_summed == CHECKSUM_NONE) { | |
3271 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
3272 | skb->csum_level = 0; | |
3273 | } | |
3274 | } | |
3275 | ||
5a212329 TH |
3276 | static inline void __skb_mark_checksum_bad(struct sk_buff *skb) |
3277 | { | |
3278 | /* Mark current checksum as bad (typically called from GRO | |
3279 | * path). In the case that ip_summed is CHECKSUM_NONE | |
3280 | * this must be the first checksum encountered in the packet. | |
3281 | * When ip_summed is CHECKSUM_UNNECESSARY, this is the first | |
3282 | * checksum after the last one validated. For UDP, a zero | |
3283 | * checksum can not be marked as bad. | |
3284 | */ | |
3285 | ||
3286 | if (skb->ip_summed == CHECKSUM_NONE || | |
3287 | skb->ip_summed == CHECKSUM_UNNECESSARY) | |
3288 | skb->csum_bad = 1; | |
3289 | } | |
3290 | ||
76ba0aae TH |
3291 | /* Check if we need to perform checksum complete validation. |
3292 | * | |
3293 | * Returns true if checksum complete is needed, false otherwise | |
3294 | * (either checksum is unnecessary or zero checksum is allowed). | |
3295 | */ | |
3296 | static inline bool __skb_checksum_validate_needed(struct sk_buff *skb, | |
3297 | bool zero_okay, | |
3298 | __sum16 check) | |
3299 | { | |
5d0c2b95 TH |
3300 | if (skb_csum_unnecessary(skb) || (zero_okay && !check)) { |
3301 | skb->csum_valid = 1; | |
77cffe23 | 3302 | __skb_decr_checksum_unnecessary(skb); |
76ba0aae TH |
3303 | return false; |
3304 | } | |
3305 | ||
3306 | return true; | |
3307 | } | |
3308 | ||
3309 | /* For small packets <= CHECKSUM_BREAK peform checksum complete directly | |
3310 | * in checksum_init. | |
3311 | */ | |
3312 | #define CHECKSUM_BREAK 76 | |
3313 | ||
4e18b9ad TH |
3314 | /* Unset checksum-complete |
3315 | * | |
3316 | * Unset checksum complete can be done when packet is being modified | |
3317 | * (uncompressed for instance) and checksum-complete value is | |
3318 | * invalidated. | |
3319 | */ | |
3320 | static inline void skb_checksum_complete_unset(struct sk_buff *skb) | |
3321 | { | |
3322 | if (skb->ip_summed == CHECKSUM_COMPLETE) | |
3323 | skb->ip_summed = CHECKSUM_NONE; | |
3324 | } | |
3325 | ||
76ba0aae TH |
3326 | /* Validate (init) checksum based on checksum complete. |
3327 | * | |
3328 | * Return values: | |
3329 | * 0: checksum is validated or try to in skb_checksum_complete. In the latter | |
3330 | * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo | |
3331 | * checksum is stored in skb->csum for use in __skb_checksum_complete | |
3332 | * non-zero: value of invalid checksum | |
3333 | * | |
3334 | */ | |
3335 | static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb, | |
3336 | bool complete, | |
3337 | __wsum psum) | |
3338 | { | |
3339 | if (skb->ip_summed == CHECKSUM_COMPLETE) { | |
3340 | if (!csum_fold(csum_add(psum, skb->csum))) { | |
5d0c2b95 | 3341 | skb->csum_valid = 1; |
76ba0aae TH |
3342 | return 0; |
3343 | } | |
5a212329 TH |
3344 | } else if (skb->csum_bad) { |
3345 | /* ip_summed == CHECKSUM_NONE in this case */ | |
c91d4606 | 3346 | return (__force __sum16)1; |
76ba0aae TH |
3347 | } |
3348 | ||
3349 | skb->csum = psum; | |
3350 | ||
5d0c2b95 TH |
3351 | if (complete || skb->len <= CHECKSUM_BREAK) { |
3352 | __sum16 csum; | |
3353 | ||
3354 | csum = __skb_checksum_complete(skb); | |
3355 | skb->csum_valid = !csum; | |
3356 | return csum; | |
3357 | } | |
76ba0aae TH |
3358 | |
3359 | return 0; | |
3360 | } | |
3361 | ||
3362 | static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto) | |
3363 | { | |
3364 | return 0; | |
3365 | } | |
3366 | ||
3367 | /* Perform checksum validate (init). Note that this is a macro since we only | |
3368 | * want to calculate the pseudo header which is an input function if necessary. | |
3369 | * First we try to validate without any computation (checksum unnecessary) and | |
3370 | * then calculate based on checksum complete calling the function to compute | |
3371 | * pseudo header. | |
3372 | * | |
3373 | * Return values: | |
3374 | * 0: checksum is validated or try to in skb_checksum_complete | |
3375 | * non-zero: value of invalid checksum | |
3376 | */ | |
3377 | #define __skb_checksum_validate(skb, proto, complete, \ | |
3378 | zero_okay, check, compute_pseudo) \ | |
3379 | ({ \ | |
3380 | __sum16 __ret = 0; \ | |
5d0c2b95 | 3381 | skb->csum_valid = 0; \ |
76ba0aae TH |
3382 | if (__skb_checksum_validate_needed(skb, zero_okay, check)) \ |
3383 | __ret = __skb_checksum_validate_complete(skb, \ | |
3384 | complete, compute_pseudo(skb, proto)); \ | |
3385 | __ret; \ | |
3386 | }) | |
3387 | ||
3388 | #define skb_checksum_init(skb, proto, compute_pseudo) \ | |
3389 | __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo) | |
3390 | ||
3391 | #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \ | |
3392 | __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo) | |
3393 | ||
3394 | #define skb_checksum_validate(skb, proto, compute_pseudo) \ | |
3395 | __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo) | |
3396 | ||
3397 | #define skb_checksum_validate_zero_check(skb, proto, check, \ | |
3398 | compute_pseudo) \ | |
096a4cfa | 3399 | __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo) |
76ba0aae TH |
3400 | |
3401 | #define skb_checksum_simple_validate(skb) \ | |
3402 | __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo) | |
3403 | ||
d96535a1 TH |
3404 | static inline bool __skb_checksum_convert_check(struct sk_buff *skb) |
3405 | { | |
3406 | return (skb->ip_summed == CHECKSUM_NONE && | |
3407 | skb->csum_valid && !skb->csum_bad); | |
3408 | } | |
3409 | ||
3410 | static inline void __skb_checksum_convert(struct sk_buff *skb, | |
3411 | __sum16 check, __wsum pseudo) | |
3412 | { | |
3413 | skb->csum = ~pseudo; | |
3414 | skb->ip_summed = CHECKSUM_COMPLETE; | |
3415 | } | |
3416 | ||
3417 | #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \ | |
3418 | do { \ | |
3419 | if (__skb_checksum_convert_check(skb)) \ | |
3420 | __skb_checksum_convert(skb, check, \ | |
3421 | compute_pseudo(skb, proto)); \ | |
3422 | } while (0) | |
3423 | ||
15e2396d TH |
3424 | static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr, |
3425 | u16 start, u16 offset) | |
3426 | { | |
3427 | skb->ip_summed = CHECKSUM_PARTIAL; | |
3428 | skb->csum_start = ((unsigned char *)ptr + start) - skb->head; | |
3429 | skb->csum_offset = offset - start; | |
3430 | } | |
3431 | ||
dcdc8994 TH |
3432 | /* Update skbuf and packet to reflect the remote checksum offload operation. |
3433 | * When called, ptr indicates the starting point for skb->csum when | |
3434 | * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete | |
3435 | * here, skb_postpull_rcsum is done so skb->csum start is ptr. | |
3436 | */ | |
3437 | static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr, | |
15e2396d | 3438 | int start, int offset, bool nopartial) |
dcdc8994 TH |
3439 | { |
3440 | __wsum delta; | |
3441 | ||
15e2396d TH |
3442 | if (!nopartial) { |
3443 | skb_remcsum_adjust_partial(skb, ptr, start, offset); | |
3444 | return; | |
3445 | } | |
3446 | ||
dcdc8994 TH |
3447 | if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) { |
3448 | __skb_checksum_complete(skb); | |
3449 | skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data); | |
3450 | } | |
3451 | ||
3452 | delta = remcsum_adjust(ptr, skb->csum, start, offset); | |
3453 | ||
3454 | /* Adjust skb->csum since we changed the packet */ | |
3455 | skb->csum = csum_add(skb->csum, delta); | |
3456 | } | |
3457 | ||
5f79e0f9 | 3458 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) |
7965bd4d | 3459 | void nf_conntrack_destroy(struct nf_conntrack *nfct); |
1da177e4 LT |
3460 | static inline void nf_conntrack_put(struct nf_conntrack *nfct) |
3461 | { | |
3462 | if (nfct && atomic_dec_and_test(&nfct->use)) | |
de6e05c4 | 3463 | nf_conntrack_destroy(nfct); |
1da177e4 LT |
3464 | } |
3465 | static inline void nf_conntrack_get(struct nf_conntrack *nfct) | |
3466 | { | |
3467 | if (nfct) | |
3468 | atomic_inc(&nfct->use); | |
3469 | } | |
2fc72c7b | 3470 | #endif |
34666d46 | 3471 | #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) |
1da177e4 LT |
3472 | static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge) |
3473 | { | |
3474 | if (nf_bridge && atomic_dec_and_test(&nf_bridge->use)) | |
3475 | kfree(nf_bridge); | |
3476 | } | |
3477 | static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge) | |
3478 | { | |
3479 | if (nf_bridge) | |
3480 | atomic_inc(&nf_bridge->use); | |
3481 | } | |
3482 | #endif /* CONFIG_BRIDGE_NETFILTER */ | |
a193a4ab PM |
3483 | static inline void nf_reset(struct sk_buff *skb) |
3484 | { | |
5f79e0f9 | 3485 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) |
a193a4ab PM |
3486 | nf_conntrack_put(skb->nfct); |
3487 | skb->nfct = NULL; | |
2fc72c7b | 3488 | #endif |
34666d46 | 3489 | #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) |
a193a4ab PM |
3490 | nf_bridge_put(skb->nf_bridge); |
3491 | skb->nf_bridge = NULL; | |
3492 | #endif | |
3493 | } | |
3494 | ||
124dff01 PM |
3495 | static inline void nf_reset_trace(struct sk_buff *skb) |
3496 | { | |
478b360a | 3497 | #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES) |
130549fe G |
3498 | skb->nf_trace = 0; |
3499 | #endif | |
a193a4ab PM |
3500 | } |
3501 | ||
edda553c | 3502 | /* Note: This doesn't put any conntrack and bridge info in dst. */ |
b1937227 ED |
3503 | static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src, |
3504 | bool copy) | |
edda553c | 3505 | { |
5f79e0f9 | 3506 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) |
edda553c YK |
3507 | dst->nfct = src->nfct; |
3508 | nf_conntrack_get(src->nfct); | |
b1937227 ED |
3509 | if (copy) |
3510 | dst->nfctinfo = src->nfctinfo; | |
2fc72c7b | 3511 | #endif |
34666d46 | 3512 | #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) |
edda553c YK |
3513 | dst->nf_bridge = src->nf_bridge; |
3514 | nf_bridge_get(src->nf_bridge); | |
3515 | #endif | |
478b360a | 3516 | #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES) |
b1937227 ED |
3517 | if (copy) |
3518 | dst->nf_trace = src->nf_trace; | |
478b360a | 3519 | #endif |
edda553c YK |
3520 | } |
3521 | ||
e7ac05f3 YK |
3522 | static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src) |
3523 | { | |
e7ac05f3 | 3524 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) |
5f79e0f9 | 3525 | nf_conntrack_put(dst->nfct); |
2fc72c7b | 3526 | #endif |
34666d46 | 3527 | #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) |
e7ac05f3 YK |
3528 | nf_bridge_put(dst->nf_bridge); |
3529 | #endif | |
b1937227 | 3530 | __nf_copy(dst, src, true); |
e7ac05f3 YK |
3531 | } |
3532 | ||
984bc16c JM |
3533 | #ifdef CONFIG_NETWORK_SECMARK |
3534 | static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) | |
3535 | { | |
3536 | to->secmark = from->secmark; | |
3537 | } | |
3538 | ||
3539 | static inline void skb_init_secmark(struct sk_buff *skb) | |
3540 | { | |
3541 | skb->secmark = 0; | |
3542 | } | |
3543 | #else | |
3544 | static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from) | |
3545 | { } | |
3546 | ||
3547 | static inline void skb_init_secmark(struct sk_buff *skb) | |
3548 | { } | |
3549 | #endif | |
3550 | ||
574f7194 EB |
3551 | static inline bool skb_irq_freeable(const struct sk_buff *skb) |
3552 | { | |
3553 | return !skb->destructor && | |
3554 | #if IS_ENABLED(CONFIG_XFRM) | |
3555 | !skb->sp && | |
3556 | #endif | |
3557 | #if IS_ENABLED(CONFIG_NF_CONNTRACK) | |
3558 | !skb->nfct && | |
3559 | #endif | |
3560 | !skb->_skb_refdst && | |
3561 | !skb_has_frag_list(skb); | |
3562 | } | |
3563 | ||
f25f4e44 PWJ |
3564 | static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping) |
3565 | { | |
f25f4e44 | 3566 | skb->queue_mapping = queue_mapping; |
f25f4e44 PWJ |
3567 | } |
3568 | ||
9247744e | 3569 | static inline u16 skb_get_queue_mapping(const struct sk_buff *skb) |
4e3ab47a | 3570 | { |
4e3ab47a | 3571 | return skb->queue_mapping; |
4e3ab47a PE |
3572 | } |
3573 | ||
f25f4e44 PWJ |
3574 | static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from) |
3575 | { | |
f25f4e44 | 3576 | to->queue_mapping = from->queue_mapping; |
f25f4e44 PWJ |
3577 | } |
3578 | ||
d5a9e24a DM |
3579 | static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue) |
3580 | { | |
3581 | skb->queue_mapping = rx_queue + 1; | |
3582 | } | |
3583 | ||
9247744e | 3584 | static inline u16 skb_get_rx_queue(const struct sk_buff *skb) |
d5a9e24a DM |
3585 | { |
3586 | return skb->queue_mapping - 1; | |
3587 | } | |
3588 | ||
9247744e | 3589 | static inline bool skb_rx_queue_recorded(const struct sk_buff *skb) |
d5a9e24a | 3590 | { |
a02cec21 | 3591 | return skb->queue_mapping != 0; |
d5a9e24a DM |
3592 | } |
3593 | ||
def8b4fa AD |
3594 | static inline struct sec_path *skb_sec_path(struct sk_buff *skb) |
3595 | { | |
0b3d8e08 | 3596 | #ifdef CONFIG_XFRM |
def8b4fa | 3597 | return skb->sp; |
def8b4fa | 3598 | #else |
def8b4fa | 3599 | return NULL; |
def8b4fa | 3600 | #endif |
0b3d8e08 | 3601 | } |
def8b4fa | 3602 | |
68c33163 PS |
3603 | /* Keeps track of mac header offset relative to skb->head. |
3604 | * It is useful for TSO of Tunneling protocol. e.g. GRE. | |
3605 | * For non-tunnel skb it points to skb_mac_header() and for | |
3347c960 ED |
3606 | * tunnel skb it points to outer mac header. |
3607 | * Keeps track of level of encapsulation of network headers. | |
3608 | */ | |
68c33163 | 3609 | struct skb_gso_cb { |
802ab55a AD |
3610 | union { |
3611 | int mac_offset; | |
3612 | int data_offset; | |
3613 | }; | |
3347c960 | 3614 | int encap_level; |
76443456 | 3615 | __wsum csum; |
7e2b10c1 | 3616 | __u16 csum_start; |
68c33163 | 3617 | }; |
9207f9d4 KK |
3618 | #define SKB_SGO_CB_OFFSET 32 |
3619 | #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET)) | |
68c33163 PS |
3620 | |
3621 | static inline int skb_tnl_header_len(const struct sk_buff *inner_skb) | |
3622 | { | |
3623 | return (skb_mac_header(inner_skb) - inner_skb->head) - | |
3624 | SKB_GSO_CB(inner_skb)->mac_offset; | |
3625 | } | |
3626 | ||
1e2bd517 PS |
3627 | static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra) |
3628 | { | |
3629 | int new_headroom, headroom; | |
3630 | int ret; | |
3631 | ||
3632 | headroom = skb_headroom(skb); | |
3633 | ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC); | |
3634 | if (ret) | |
3635 | return ret; | |
3636 | ||
3637 | new_headroom = skb_headroom(skb); | |
3638 | SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom); | |
3639 | return 0; | |
3640 | } | |
3641 | ||
08b64fcc AD |
3642 | static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res) |
3643 | { | |
3644 | /* Do not update partial checksums if remote checksum is enabled. */ | |
3645 | if (skb->remcsum_offload) | |
3646 | return; | |
3647 | ||
3648 | SKB_GSO_CB(skb)->csum = res; | |
3649 | SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head; | |
3650 | } | |
3651 | ||
7e2b10c1 TH |
3652 | /* Compute the checksum for a gso segment. First compute the checksum value |
3653 | * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and | |
3654 | * then add in skb->csum (checksum from csum_start to end of packet). | |
3655 | * skb->csum and csum_start are then updated to reflect the checksum of the | |
3656 | * resultant packet starting from the transport header-- the resultant checksum | |
3657 | * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo | |
3658 | * header. | |
3659 | */ | |
3660 | static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res) | |
3661 | { | |
76443456 AD |
3662 | unsigned char *csum_start = skb_transport_header(skb); |
3663 | int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start; | |
3664 | __wsum partial = SKB_GSO_CB(skb)->csum; | |
7e2b10c1 | 3665 | |
76443456 AD |
3666 | SKB_GSO_CB(skb)->csum = res; |
3667 | SKB_GSO_CB(skb)->csum_start = csum_start - skb->head; | |
7e2b10c1 | 3668 | |
76443456 | 3669 | return csum_fold(csum_partial(csum_start, plen, partial)); |
7e2b10c1 TH |
3670 | } |
3671 | ||
bdcc0924 | 3672 | static inline bool skb_is_gso(const struct sk_buff *skb) |
89114afd HX |
3673 | { |
3674 | return skb_shinfo(skb)->gso_size; | |
3675 | } | |
3676 | ||
36a8f39e | 3677 | /* Note: Should be called only if skb_is_gso(skb) is true */ |
bdcc0924 | 3678 | static inline bool skb_is_gso_v6(const struct sk_buff *skb) |
eabd7e35 BG |
3679 | { |
3680 | return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6; | |
3681 | } | |
3682 | ||
7965bd4d | 3683 | void __skb_warn_lro_forwarding(const struct sk_buff *skb); |
4497b076 BH |
3684 | |
3685 | static inline bool skb_warn_if_lro(const struct sk_buff *skb) | |
3686 | { | |
3687 | /* LRO sets gso_size but not gso_type, whereas if GSO is really | |
3688 | * wanted then gso_type will be set. */ | |
05bdd2f1 ED |
3689 | const struct skb_shared_info *shinfo = skb_shinfo(skb); |
3690 | ||
b78462eb AD |
3691 | if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 && |
3692 | unlikely(shinfo->gso_type == 0)) { | |
4497b076 BH |
3693 | __skb_warn_lro_forwarding(skb); |
3694 | return true; | |
3695 | } | |
3696 | return false; | |
3697 | } | |
3698 | ||
35fc92a9 HX |
3699 | static inline void skb_forward_csum(struct sk_buff *skb) |
3700 | { | |
3701 | /* Unfortunately we don't support this one. Any brave souls? */ | |
3702 | if (skb->ip_summed == CHECKSUM_COMPLETE) | |
3703 | skb->ip_summed = CHECKSUM_NONE; | |
3704 | } | |
3705 | ||
bc8acf2c ED |
3706 | /** |
3707 | * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE | |
3708 | * @skb: skb to check | |
3709 | * | |
3710 | * fresh skbs have their ip_summed set to CHECKSUM_NONE. | |
3711 | * Instead of forcing ip_summed to CHECKSUM_NONE, we can | |
3712 | * use this helper, to document places where we make this assertion. | |
3713 | */ | |
05bdd2f1 | 3714 | static inline void skb_checksum_none_assert(const struct sk_buff *skb) |
bc8acf2c ED |
3715 | { |
3716 | #ifdef DEBUG | |
3717 | BUG_ON(skb->ip_summed != CHECKSUM_NONE); | |
3718 | #endif | |
3719 | } | |
3720 | ||
f35d9d8a | 3721 | bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off); |
a6686f2f | 3722 | |
ed1f50c3 | 3723 | int skb_checksum_setup(struct sk_buff *skb, bool recalculate); |
9afd85c9 LL |
3724 | struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb, |
3725 | unsigned int transport_len, | |
3726 | __sum16(*skb_chkf)(struct sk_buff *skb)); | |
ed1f50c3 | 3727 | |
3a7c1ee4 AD |
3728 | /** |
3729 | * skb_head_is_locked - Determine if the skb->head is locked down | |
3730 | * @skb: skb to check | |
3731 | * | |
3732 | * The head on skbs build around a head frag can be removed if they are | |
3733 | * not cloned. This function returns true if the skb head is locked down | |
3734 | * due to either being allocated via kmalloc, or by being a clone with | |
3735 | * multiple references to the head. | |
3736 | */ | |
3737 | static inline bool skb_head_is_locked(const struct sk_buff *skb) | |
3738 | { | |
3739 | return !skb->head_frag || skb_cloned(skb); | |
3740 | } | |
fe6cc55f FW |
3741 | |
3742 | /** | |
3743 | * skb_gso_network_seglen - Return length of individual segments of a gso packet | |
3744 | * | |
3745 | * @skb: GSO skb | |
3746 | * | |
3747 | * skb_gso_network_seglen is used to determine the real size of the | |
3748 | * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP). | |
3749 | * | |
3750 | * The MAC/L2 header is not accounted for. | |
3751 | */ | |
3752 | static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb) | |
3753 | { | |
3754 | unsigned int hdr_len = skb_transport_header(skb) - | |
3755 | skb_network_header(skb); | |
3756 | return hdr_len + skb_gso_transport_seglen(skb); | |
3757 | } | |
ee122c79 | 3758 | |
179bc67f EC |
3759 | /* Local Checksum Offload. |
3760 | * Compute outer checksum based on the assumption that the | |
3761 | * inner checksum will be offloaded later. | |
e8ae7b00 EC |
3762 | * See Documentation/networking/checksum-offloads.txt for |
3763 | * explanation of how this works. | |
179bc67f EC |
3764 | * Fill in outer checksum adjustment (e.g. with sum of outer |
3765 | * pseudo-header) before calling. | |
3766 | * Also ensure that inner checksum is in linear data area. | |
3767 | */ | |
3768 | static inline __wsum lco_csum(struct sk_buff *skb) | |
3769 | { | |
9e74a6da AD |
3770 | unsigned char *csum_start = skb_checksum_start(skb); |
3771 | unsigned char *l4_hdr = skb_transport_header(skb); | |
3772 | __wsum partial; | |
179bc67f EC |
3773 | |
3774 | /* Start with complement of inner checksum adjustment */ | |
9e74a6da AD |
3775 | partial = ~csum_unfold(*(__force __sum16 *)(csum_start + |
3776 | skb->csum_offset)); | |
3777 | ||
179bc67f | 3778 | /* Add in checksum of our headers (incl. outer checksum |
9e74a6da | 3779 | * adjustment filled in by caller) and return result. |
179bc67f | 3780 | */ |
9e74a6da | 3781 | return csum_partial(l4_hdr, csum_start - l4_hdr, partial); |
179bc67f EC |
3782 | } |
3783 | ||
1da177e4 LT |
3784 | #endif /* __KERNEL__ */ |
3785 | #endif /* _LINUX_SKBUFF_H */ |