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> | |
21 | #include <linux/cache.h> | |
22 | ||
23 | #include <asm/atomic.h> | |
24 | #include <asm/types.h> | |
25 | #include <linux/spinlock.h> | |
1da177e4 | 26 | #include <linux/net.h> |
3fc7e8a6 | 27 | #include <linux/textsearch.h> |
1da177e4 | 28 | #include <net/checksum.h> |
a80958f4 | 29 | #include <linux/rcupdate.h> |
97fc2f08 | 30 | #include <linux/dmaengine.h> |
b7aa0bf7 | 31 | #include <linux/hrtimer.h> |
1da177e4 | 32 | |
60476372 | 33 | /* Don't change this without changing skb_csum_unnecessary! */ |
1da177e4 | 34 | #define CHECKSUM_NONE 0 |
60476372 HX |
35 | #define CHECKSUM_UNNECESSARY 1 |
36 | #define CHECKSUM_COMPLETE 2 | |
37 | #define CHECKSUM_PARTIAL 3 | |
1da177e4 LT |
38 | |
39 | #define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \ | |
40 | ~(SMP_CACHE_BYTES - 1)) | |
fc910a27 | 41 | #define SKB_WITH_OVERHEAD(X) \ |
deea84b0 | 42 | ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info))) |
fc910a27 DM |
43 | #define SKB_MAX_ORDER(X, ORDER) \ |
44 | SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X)) | |
1da177e4 LT |
45 | #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0)) |
46 | #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2)) | |
47 | ||
48 | /* A. Checksumming of received packets by device. | |
49 | * | |
50 | * NONE: device failed to checksum this packet. | |
51 | * skb->csum is undefined. | |
52 | * | |
53 | * UNNECESSARY: device parsed packet and wouldbe verified checksum. | |
54 | * skb->csum is undefined. | |
55 | * It is bad option, but, unfortunately, many of vendors do this. | |
56 | * Apparently with secret goal to sell you new device, when you | |
57 | * will add new protocol to your host. F.e. IPv6. 8) | |
58 | * | |
84fa7933 | 59 | * COMPLETE: the most generic way. Device supplied checksum of _all_ |
1da177e4 LT |
60 | * the packet as seen by netif_rx in skb->csum. |
61 | * NOTE: Even if device supports only some protocols, but | |
84fa7933 | 62 | * is able to produce some skb->csum, it MUST use COMPLETE, |
1da177e4 LT |
63 | * not UNNECESSARY. |
64 | * | |
c6c6e3e0 HX |
65 | * PARTIAL: identical to the case for output below. This may occur |
66 | * on a packet received directly from another Linux OS, e.g., | |
67 | * a virtualised Linux kernel on the same host. The packet can | |
68 | * be treated in the same way as UNNECESSARY except that on | |
69 | * output (i.e., forwarding) the checksum must be filled in | |
70 | * by the OS or the hardware. | |
71 | * | |
1da177e4 LT |
72 | * B. Checksumming on output. |
73 | * | |
74 | * NONE: skb is checksummed by protocol or csum is not required. | |
75 | * | |
84fa7933 | 76 | * PARTIAL: device is required to csum packet as seen by hard_start_xmit |
c6c6e3e0 HX |
77 | * from skb->csum_start to the end and to record the checksum |
78 | * at skb->csum_start + skb->csum_offset. | |
1da177e4 LT |
79 | * |
80 | * Device must show its capabilities in dev->features, set | |
81 | * at device setup time. | |
82 | * NETIF_F_HW_CSUM - it is clever device, it is able to checksum | |
83 | * everything. | |
84 | * NETIF_F_NO_CSUM - loopback or reliable single hop media. | |
85 | * NETIF_F_IP_CSUM - device is dumb. It is able to csum only | |
86 | * TCP/UDP over IPv4. Sigh. Vendors like this | |
87 | * way by an unknown reason. Though, see comment above | |
88 | * about CHECKSUM_UNNECESSARY. 8) | |
c6c6e3e0 | 89 | * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead. |
1da177e4 LT |
90 | * |
91 | * Any questions? No questions, good. --ANK | |
92 | */ | |
93 | ||
1da177e4 | 94 | struct net_device; |
716ea3a7 | 95 | struct scatterlist; |
9c55e01c | 96 | struct pipe_inode_info; |
1da177e4 | 97 | |
5f79e0f9 | 98 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) |
1da177e4 LT |
99 | struct nf_conntrack { |
100 | atomic_t use; | |
1da177e4 | 101 | }; |
5f79e0f9 | 102 | #endif |
1da177e4 LT |
103 | |
104 | #ifdef CONFIG_BRIDGE_NETFILTER | |
105 | struct nf_bridge_info { | |
106 | atomic_t use; | |
107 | struct net_device *physindev; | |
108 | struct net_device *physoutdev; | |
1da177e4 LT |
109 | unsigned int mask; |
110 | unsigned long data[32 / sizeof(unsigned long)]; | |
111 | }; | |
112 | #endif | |
113 | ||
1da177e4 LT |
114 | struct sk_buff_head { |
115 | /* These two members must be first. */ | |
116 | struct sk_buff *next; | |
117 | struct sk_buff *prev; | |
118 | ||
119 | __u32 qlen; | |
120 | spinlock_t lock; | |
121 | }; | |
122 | ||
123 | struct sk_buff; | |
124 | ||
125 | /* To allow 64K frame to be packed as single skb without frag_list */ | |
126 | #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2) | |
127 | ||
128 | typedef struct skb_frag_struct skb_frag_t; | |
129 | ||
130 | struct skb_frag_struct { | |
131 | struct page *page; | |
a309bb07 DM |
132 | __u32 page_offset; |
133 | __u32 size; | |
1da177e4 LT |
134 | }; |
135 | ||
ac45f602 PO |
136 | #define HAVE_HW_TIME_STAMP |
137 | ||
138 | /** | |
d3a21be8 | 139 | * struct skb_shared_hwtstamps - hardware time stamps |
ac45f602 PO |
140 | * @hwtstamp: hardware time stamp transformed into duration |
141 | * since arbitrary point in time | |
142 | * @syststamp: hwtstamp transformed to system time base | |
143 | * | |
144 | * Software time stamps generated by ktime_get_real() are stored in | |
145 | * skb->tstamp. The relation between the different kinds of time | |
146 | * stamps is as follows: | |
147 | * | |
148 | * syststamp and tstamp can be compared against each other in | |
149 | * arbitrary combinations. The accuracy of a | |
150 | * syststamp/tstamp/"syststamp from other device" comparison is | |
151 | * limited by the accuracy of the transformation into system time | |
152 | * base. This depends on the device driver and its underlying | |
153 | * hardware. | |
154 | * | |
155 | * hwtstamps can only be compared against other hwtstamps from | |
156 | * the same device. | |
157 | * | |
158 | * This structure is attached to packets as part of the | |
159 | * &skb_shared_info. Use skb_hwtstamps() to get a pointer. | |
160 | */ | |
161 | struct skb_shared_hwtstamps { | |
162 | ktime_t hwtstamp; | |
163 | ktime_t syststamp; | |
164 | }; | |
165 | ||
166 | /** | |
d3a21be8 | 167 | * struct skb_shared_tx - instructions for time stamping of outgoing packets |
ac45f602 PO |
168 | * @hardware: generate hardware time stamp |
169 | * @software: generate software time stamp | |
170 | * @in_progress: device driver is going to provide | |
171 | * hardware time stamp | |
4b21cd4e | 172 | * @flags: all shared_tx flags |
ac45f602 PO |
173 | * |
174 | * These flags are attached to packets as part of the | |
175 | * &skb_shared_info. Use skb_tx() to get a pointer. | |
176 | */ | |
177 | union skb_shared_tx { | |
178 | struct { | |
179 | __u8 hardware:1, | |
180 | software:1, | |
181 | in_progress:1; | |
182 | }; | |
183 | __u8 flags; | |
184 | }; | |
185 | ||
1da177e4 LT |
186 | /* This data is invariant across clones and lives at |
187 | * the end of the header data, ie. at skb->end. | |
188 | */ | |
189 | struct skb_shared_info { | |
190 | atomic_t dataref; | |
4947d3ef | 191 | unsigned short nr_frags; |
7967168c | 192 | unsigned short gso_size; |
042a53a9 ED |
193 | #ifdef CONFIG_HAS_DMA |
194 | dma_addr_t dma_head; | |
195 | #endif | |
7967168c HX |
196 | /* Warning: this field is not always filled in (UFO)! */ |
197 | unsigned short gso_segs; | |
198 | unsigned short gso_type; | |
ae08e1f0 | 199 | __be32 ip6_frag_id; |
ac45f602 | 200 | union skb_shared_tx tx_flags; |
1da177e4 | 201 | struct sk_buff *frag_list; |
ac45f602 | 202 | struct skb_shared_hwtstamps hwtstamps; |
1da177e4 | 203 | skb_frag_t frags[MAX_SKB_FRAGS]; |
271bff7a | 204 | #ifdef CONFIG_HAS_DMA |
042a53a9 | 205 | dma_addr_t dma_maps[MAX_SKB_FRAGS]; |
271bff7a | 206 | #endif |
69e3c75f JB |
207 | /* Intermediate layers must ensure that destructor_arg |
208 | * remains valid until skb destructor */ | |
209 | void * destructor_arg; | |
1da177e4 LT |
210 | }; |
211 | ||
212 | /* We divide dataref into two halves. The higher 16 bits hold references | |
213 | * to the payload part of skb->data. The lower 16 bits hold references to | |
334a8132 PM |
214 | * the entire skb->data. A clone of a headerless skb holds the length of |
215 | * the header in skb->hdr_len. | |
1da177e4 LT |
216 | * |
217 | * All users must obey the rule that the skb->data reference count must be | |
218 | * greater than or equal to the payload reference count. | |
219 | * | |
220 | * Holding a reference to the payload part means that the user does not | |
221 | * care about modifications to the header part of skb->data. | |
222 | */ | |
223 | #define SKB_DATAREF_SHIFT 16 | |
224 | #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1) | |
225 | ||
d179cd12 DM |
226 | |
227 | enum { | |
228 | SKB_FCLONE_UNAVAILABLE, | |
229 | SKB_FCLONE_ORIG, | |
230 | SKB_FCLONE_CLONE, | |
231 | }; | |
232 | ||
7967168c HX |
233 | enum { |
234 | SKB_GSO_TCPV4 = 1 << 0, | |
f83ef8c0 | 235 | SKB_GSO_UDP = 1 << 1, |
576a30eb HX |
236 | |
237 | /* This indicates the skb is from an untrusted source. */ | |
238 | SKB_GSO_DODGY = 1 << 2, | |
b0da8537 MC |
239 | |
240 | /* This indicates the tcp segment has CWR set. */ | |
f83ef8c0 HX |
241 | SKB_GSO_TCP_ECN = 1 << 3, |
242 | ||
243 | SKB_GSO_TCPV6 = 1 << 4, | |
01d5b2fc CL |
244 | |
245 | SKB_GSO_FCOE = 1 << 5, | |
7967168c HX |
246 | }; |
247 | ||
2e07fa9c ACM |
248 | #if BITS_PER_LONG > 32 |
249 | #define NET_SKBUFF_DATA_USES_OFFSET 1 | |
250 | #endif | |
251 | ||
252 | #ifdef NET_SKBUFF_DATA_USES_OFFSET | |
253 | typedef unsigned int sk_buff_data_t; | |
254 | #else | |
255 | typedef unsigned char *sk_buff_data_t; | |
256 | #endif | |
257 | ||
1da177e4 LT |
258 | /** |
259 | * struct sk_buff - socket buffer | |
260 | * @next: Next buffer in list | |
261 | * @prev: Previous buffer in list | |
1da177e4 | 262 | * @sk: Socket we are owned by |
325ed823 | 263 | * @tstamp: Time we arrived |
1da177e4 | 264 | * @dev: Device we arrived on/are leaving by |
be52178b | 265 | * @transport_header: Transport layer header |
b0e380b1 ACM |
266 | * @network_header: Network layer header |
267 | * @mac_header: Link layer header | |
a42fc8f6 | 268 | * @_skb_dst: destination entry |
67be2dd1 | 269 | * @sp: the security path, used for xfrm |
1da177e4 LT |
270 | * @cb: Control buffer. Free for use by every layer. Put private vars here |
271 | * @len: Length of actual data | |
272 | * @data_len: Data length | |
273 | * @mac_len: Length of link layer header | |
334a8132 | 274 | * @hdr_len: writable header length of cloned skb |
663ead3b HX |
275 | * @csum: Checksum (must include start/offset pair) |
276 | * @csum_start: Offset from skb->head where checksumming should start | |
277 | * @csum_offset: Offset from csum_start where checksum should be stored | |
67be2dd1 | 278 | * @local_df: allow local fragmentation |
1da177e4 LT |
279 | * @cloned: Head may be cloned (check refcnt to be sure) |
280 | * @nohdr: Payload reference only, must not modify header | |
281 | * @pkt_type: Packet class | |
c83c2486 | 282 | * @fclone: skbuff clone status |
1da177e4 LT |
283 | * @ip_summed: Driver fed us an IP checksum |
284 | * @priority: Packet queueing priority | |
285 | * @users: User count - see {datagram,tcp}.c | |
286 | * @protocol: Packet protocol from driver | |
1da177e4 LT |
287 | * @truesize: Buffer size |
288 | * @head: Head of buffer | |
289 | * @data: Data head pointer | |
290 | * @tail: Tail pointer | |
291 | * @end: End pointer | |
292 | * @destructor: Destruct function | |
82e91ffe | 293 | * @mark: Generic packet mark |
1da177e4 | 294 | * @nfct: Associated connection, if any |
c83c2486 | 295 | * @ipvs_property: skbuff is owned by ipvs |
31729363 RD |
296 | * @peeked: this packet has been seen already, so stats have been |
297 | * done for it, don't do them again | |
ba9dda3a | 298 | * @nf_trace: netfilter packet trace flag |
1da177e4 | 299 | * @nfctinfo: Relationship of this skb to the connection |
461ddf3b | 300 | * @nfct_reasm: netfilter conntrack re-assembly pointer |
1da177e4 | 301 | * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c |
f25f4e44 PWJ |
302 | * @iif: ifindex of device we arrived on |
303 | * @queue_mapping: Queue mapping for multiqueue devices | |
1da177e4 LT |
304 | * @tc_index: Traffic control index |
305 | * @tc_verd: traffic control verdict | |
553a5672 | 306 | * @ndisc_nodetype: router type (from link layer) |
f4b8ea78 RD |
307 | * @dma_cookie: a cookie to one of several possible DMA operations |
308 | * done by skb DMA functions | |
984bc16c | 309 | * @secmark: security marking |
6aa895b0 | 310 | * @vlan_tci: vlan tag control information |
1da177e4 LT |
311 | */ |
312 | ||
313 | struct sk_buff { | |
314 | /* These two members must be first. */ | |
315 | struct sk_buff *next; | |
316 | struct sk_buff *prev; | |
317 | ||
1da177e4 | 318 | struct sock *sk; |
b7aa0bf7 | 319 | ktime_t tstamp; |
1da177e4 | 320 | struct net_device *dev; |
1da177e4 | 321 | |
e5b9215e | 322 | unsigned long _skb_dst; |
def8b4fa | 323 | #ifdef CONFIG_XFRM |
1da177e4 | 324 | struct sec_path *sp; |
def8b4fa | 325 | #endif |
1da177e4 LT |
326 | /* |
327 | * This is the control buffer. It is free to use for every | |
328 | * layer. Please put your private variables there. If you | |
329 | * want to keep them across layers you have to do a skb_clone() | |
330 | * first. This is owned by whoever has the skb queued ATM. | |
331 | */ | |
3e3850e9 | 332 | char cb[48]; |
1da177e4 LT |
333 | |
334 | unsigned int len, | |
334a8132 PM |
335 | data_len; |
336 | __u16 mac_len, | |
337 | hdr_len; | |
ff1dcadb AV |
338 | union { |
339 | __wsum csum; | |
663ead3b HX |
340 | struct { |
341 | __u16 csum_start; | |
342 | __u16 csum_offset; | |
343 | }; | |
ff1dcadb | 344 | }; |
1da177e4 | 345 | __u32 priority; |
fe55f6d5 | 346 | kmemcheck_bitfield_begin(flags1); |
1cbb3380 TG |
347 | __u8 local_df:1, |
348 | cloned:1, | |
349 | ip_summed:2, | |
6869c4d8 HW |
350 | nohdr:1, |
351 | nfctinfo:3; | |
d179cd12 | 352 | __u8 pkt_type:3, |
b84f4cc9 | 353 | fclone:2, |
ba9dda3a | 354 | ipvs_property:1, |
a59322be | 355 | peeked:1, |
ba9dda3a | 356 | nf_trace:1; |
fe55f6d5 | 357 | kmemcheck_bitfield_end(flags1); |
a0d3bea3 | 358 | __be16 protocol; |
1da177e4 LT |
359 | |
360 | void (*destructor)(struct sk_buff *skb); | |
9fb9cbb1 | 361 | #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE) |
5f79e0f9 | 362 | struct nf_conntrack *nfct; |
9fb9cbb1 YK |
363 | struct sk_buff *nfct_reasm; |
364 | #endif | |
1da177e4 LT |
365 | #ifdef CONFIG_BRIDGE_NETFILTER |
366 | struct nf_bridge_info *nf_bridge; | |
367 | #endif | |
f25f4e44 PWJ |
368 | |
369 | int iif; | |
370 | __u16 queue_mapping; | |
1da177e4 | 371 | #ifdef CONFIG_NET_SCHED |
b6b99eb5 | 372 | __u16 tc_index; /* traffic control index */ |
1da177e4 | 373 | #ifdef CONFIG_NET_CLS_ACT |
b6b99eb5 | 374 | __u16 tc_verd; /* traffic control verdict */ |
1da177e4 | 375 | #endif |
1da177e4 | 376 | #endif |
fe55f6d5 VN |
377 | |
378 | kmemcheck_bitfield_begin(flags2); | |
de357cc0 | 379 | #ifdef CONFIG_IPV6_NDISC_NODETYPE |
fadf6bf0 | 380 | __u8 ndisc_nodetype:2; |
d0f09804 | 381 | #endif |
fe55f6d5 VN |
382 | kmemcheck_bitfield_end(flags2); |
383 | ||
72bce627 | 384 | /* 0/14 bit hole */ |
f25f4e44 | 385 | |
97fc2f08 CL |
386 | #ifdef CONFIG_NET_DMA |
387 | dma_cookie_t dma_cookie; | |
388 | #endif | |
984bc16c JM |
389 | #ifdef CONFIG_NETWORK_SECMARK |
390 | __u32 secmark; | |
391 | #endif | |
3b885787 NH |
392 | union { |
393 | __u32 mark; | |
394 | __u32 dropcount; | |
395 | }; | |
1da177e4 | 396 | |
6aa895b0 PM |
397 | __u16 vlan_tci; |
398 | ||
27a884dc ACM |
399 | sk_buff_data_t transport_header; |
400 | sk_buff_data_t network_header; | |
401 | sk_buff_data_t mac_header; | |
1da177e4 | 402 | /* These elements must be at the end, see alloc_skb() for details. */ |
27a884dc | 403 | sk_buff_data_t tail; |
4305b541 | 404 | sk_buff_data_t end; |
1da177e4 | 405 | unsigned char *head, |
4305b541 | 406 | *data; |
27a884dc ACM |
407 | unsigned int truesize; |
408 | atomic_t users; | |
1da177e4 LT |
409 | }; |
410 | ||
411 | #ifdef __KERNEL__ | |
412 | /* | |
413 | * Handling routines are only of interest to the kernel | |
414 | */ | |
415 | #include <linux/slab.h> | |
416 | ||
417 | #include <asm/system.h> | |
418 | ||
a40c24a1 DM |
419 | #ifdef CONFIG_HAS_DMA |
420 | #include <linux/dma-mapping.h> | |
421 | extern int skb_dma_map(struct device *dev, struct sk_buff *skb, | |
422 | enum dma_data_direction dir); | |
423 | extern void skb_dma_unmap(struct device *dev, struct sk_buff *skb, | |
424 | enum dma_data_direction dir); | |
425 | #endif | |
426 | ||
adf30907 ED |
427 | static inline struct dst_entry *skb_dst(const struct sk_buff *skb) |
428 | { | |
429 | return (struct dst_entry *)skb->_skb_dst; | |
430 | } | |
431 | ||
432 | static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst) | |
433 | { | |
434 | skb->_skb_dst = (unsigned long)dst; | |
435 | } | |
436 | ||
511c3f92 ED |
437 | static inline struct rtable *skb_rtable(const struct sk_buff *skb) |
438 | { | |
adf30907 | 439 | return (struct rtable *)skb_dst(skb); |
511c3f92 ED |
440 | } |
441 | ||