]> git.ipfire.org Git - thirdparty/qemu.git/blob - include/block/nbd.h
6d05983a55f0cfa2be94dd0b640e0e5fbbdf59e6
[thirdparty/qemu.git] / include / block / nbd.h
1 /*
2 * Copyright (C) 2016-2019 Red Hat, Inc.
3 * Copyright (C) 2005 Anthony Liguori <anthony@codemonkey.ws>
4 *
5 * Network Block Device
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; under version 2 of the License.
10 *
11 * This program is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
15 *
16 * You should have received a copy of the GNU General Public License
17 * along with this program; if not, see <http://www.gnu.org/licenses/>.
18 */
19
20 #ifndef NBD_H
21 #define NBD_H
22
23 #include "qapi/qapi-types-block.h"
24 #include "io/channel-socket.h"
25 #include "crypto/tlscreds.h"
26 #include "qapi/error.h"
27
28 /* Handshake phase structs - this struct is passed on the wire */
29
30 struct NBDOption {
31 uint64_t magic; /* NBD_OPTS_MAGIC */
32 uint32_t option; /* NBD_OPT_* */
33 uint32_t length;
34 } QEMU_PACKED;
35 typedef struct NBDOption NBDOption;
36
37 struct NBDOptionReply {
38 uint64_t magic; /* NBD_REP_MAGIC */
39 uint32_t option; /* NBD_OPT_* */
40 uint32_t type; /* NBD_REP_* */
41 uint32_t length;
42 } QEMU_PACKED;
43 typedef struct NBDOptionReply NBDOptionReply;
44
45 typedef struct NBDOptionReplyMetaContext {
46 NBDOptionReply h; /* h.type = NBD_REP_META_CONTEXT, h.length > 4 */
47 uint32_t context_id;
48 /* meta context name follows */
49 } QEMU_PACKED NBDOptionReplyMetaContext;
50
51 /* Transmission phase structs
52 *
53 * Note: these are _NOT_ the same as the network representation of an NBD
54 * request and reply!
55 */
56 struct NBDRequest {
57 uint64_t handle;
58 uint64_t from;
59 uint32_t len;
60 uint16_t flags; /* NBD_CMD_FLAG_* */
61 uint16_t type; /* NBD_CMD_* */
62 };
63 typedef struct NBDRequest NBDRequest;
64
65 typedef struct NBDSimpleReply {
66 uint32_t magic; /* NBD_SIMPLE_REPLY_MAGIC */
67 uint32_t error;
68 uint64_t handle;
69 } QEMU_PACKED NBDSimpleReply;
70
71 /* Header of all structured replies */
72 typedef struct NBDStructuredReplyChunk {
73 uint32_t magic; /* NBD_STRUCTURED_REPLY_MAGIC */
74 uint16_t flags; /* combination of NBD_REPLY_FLAG_* */
75 uint16_t type; /* NBD_REPLY_TYPE_* */
76 uint64_t handle; /* request handle */
77 uint32_t length; /* length of payload */
78 } QEMU_PACKED NBDStructuredReplyChunk;
79
80 typedef union NBDReply {
81 NBDSimpleReply simple;
82 NBDStructuredReplyChunk structured;
83 struct {
84 /* @magic and @handle fields have the same offset and size both in
85 * simple reply and structured reply chunk, so let them be accessible
86 * without ".simple." or ".structured." specification
87 */
88 uint32_t magic;
89 uint32_t _skip;
90 uint64_t handle;
91 } QEMU_PACKED;
92 } NBDReply;
93
94 /* Header of chunk for NBD_REPLY_TYPE_OFFSET_DATA */
95 typedef struct NBDStructuredReadData {
96 NBDStructuredReplyChunk h; /* h.length >= 9 */
97 uint64_t offset;
98 /* At least one byte of data payload follows, calculated from h.length */
99 } QEMU_PACKED NBDStructuredReadData;
100
101 /* Complete chunk for NBD_REPLY_TYPE_OFFSET_HOLE */
102 typedef struct NBDStructuredReadHole {
103 NBDStructuredReplyChunk h; /* h.length == 12 */
104 uint64_t offset;
105 uint32_t length;
106 } QEMU_PACKED NBDStructuredReadHole;
107
108 /* Header of all NBD_REPLY_TYPE_ERROR* errors */
109 typedef struct NBDStructuredError {
110 NBDStructuredReplyChunk h; /* h.length >= 6 */
111 uint32_t error;
112 uint16_t message_length;
113 } QEMU_PACKED NBDStructuredError;
114
115 /* Header of NBD_REPLY_TYPE_BLOCK_STATUS */
116 typedef struct NBDStructuredMeta {
117 NBDStructuredReplyChunk h; /* h.length >= 12 (at least one extent) */
118 uint32_t context_id;
119 /* extents follows */
120 } QEMU_PACKED NBDStructuredMeta;
121
122 /* Extent chunk for NBD_REPLY_TYPE_BLOCK_STATUS */
123 typedef struct NBDExtent {
124 uint32_t length;
125 uint32_t flags; /* NBD_STATE_* */
126 } QEMU_PACKED NBDExtent;
127
128 /* Transmission (export) flags: sent from server to client during handshake,
129 but describe what will happen during transmission */
130 #define NBD_FLAG_HAS_FLAGS (1 << 0) /* Flags are there */
131 #define NBD_FLAG_READ_ONLY (1 << 1) /* Device is read-only */
132 #define NBD_FLAG_SEND_FLUSH (1 << 2) /* Send FLUSH */
133 #define NBD_FLAG_SEND_FUA (1 << 3) /* Send FUA (Force Unit Access) */
134 #define NBD_FLAG_ROTATIONAL (1 << 4) /* Use elevator algorithm -
135 rotational media */
136 #define NBD_FLAG_SEND_TRIM (1 << 5) /* Send TRIM (discard) */
137 #define NBD_FLAG_SEND_WRITE_ZEROES (1 << 6) /* Send WRITE_ZEROES */
138 #define NBD_FLAG_SEND_DF (1 << 7) /* Send DF (Do not Fragment) */
139 #define NBD_FLAG_CAN_MULTI_CONN (1 << 8) /* Multi-client cache consistent */
140 #define NBD_FLAG_SEND_RESIZE (1 << 9) /* Send resize */
141 #define NBD_FLAG_SEND_CACHE (1 << 10) /* Send CACHE (prefetch) */
142
143 /* New-style handshake (global) flags, sent from server to client, and
144 control what will happen during handshake phase. */
145 #define NBD_FLAG_FIXED_NEWSTYLE (1 << 0) /* Fixed newstyle protocol. */
146 #define NBD_FLAG_NO_ZEROES (1 << 1) /* End handshake without zeroes. */
147
148 /* New-style client flags, sent from client to server to control what happens
149 during handshake phase. */
150 #define NBD_FLAG_C_FIXED_NEWSTYLE (1 << 0) /* Fixed newstyle protocol. */
151 #define NBD_FLAG_C_NO_ZEROES (1 << 1) /* End handshake without zeroes. */
152
153 /* Option requests. */
154 #define NBD_OPT_EXPORT_NAME (1)
155 #define NBD_OPT_ABORT (2)
156 #define NBD_OPT_LIST (3)
157 /* #define NBD_OPT_PEEK_EXPORT (4) not in use */
158 #define NBD_OPT_STARTTLS (5)
159 #define NBD_OPT_INFO (6)
160 #define NBD_OPT_GO (7)
161 #define NBD_OPT_STRUCTURED_REPLY (8)
162 #define NBD_OPT_LIST_META_CONTEXT (9)
163 #define NBD_OPT_SET_META_CONTEXT (10)
164
165 /* Option reply types. */
166 #define NBD_REP_ERR(value) ((UINT32_C(1) << 31) | (value))
167
168 #define NBD_REP_ACK (1) /* Data sending finished. */
169 #define NBD_REP_SERVER (2) /* Export description. */
170 #define NBD_REP_INFO (3) /* NBD_OPT_INFO/GO. */
171 #define NBD_REP_META_CONTEXT (4) /* NBD_OPT_{LIST,SET}_META_CONTEXT */
172
173 #define NBD_REP_ERR_UNSUP NBD_REP_ERR(1) /* Unknown option */
174 #define NBD_REP_ERR_POLICY NBD_REP_ERR(2) /* Server denied */
175 #define NBD_REP_ERR_INVALID NBD_REP_ERR(3) /* Invalid length */
176 #define NBD_REP_ERR_PLATFORM NBD_REP_ERR(4) /* Not compiled in */
177 #define NBD_REP_ERR_TLS_REQD NBD_REP_ERR(5) /* TLS required */
178 #define NBD_REP_ERR_UNKNOWN NBD_REP_ERR(6) /* Export unknown */
179 #define NBD_REP_ERR_SHUTDOWN NBD_REP_ERR(7) /* Server shutting down */
180 #define NBD_REP_ERR_BLOCK_SIZE_REQD NBD_REP_ERR(8) /* Need INFO_BLOCK_SIZE */
181
182 /* Info types, used during NBD_REP_INFO */
183 #define NBD_INFO_EXPORT 0
184 #define NBD_INFO_NAME 1
185 #define NBD_INFO_DESCRIPTION 2
186 #define NBD_INFO_BLOCK_SIZE 3
187
188 /* Request flags, sent from client to server during transmission phase */
189 #define NBD_CMD_FLAG_FUA (1 << 0) /* 'force unit access' during write */
190 #define NBD_CMD_FLAG_NO_HOLE (1 << 1) /* don't punch hole on zero run */
191 #define NBD_CMD_FLAG_DF (1 << 2) /* don't fragment structured read */
192 #define NBD_CMD_FLAG_REQ_ONE (1 << 3) /* only one extent in BLOCK_STATUS
193 * reply chunk */
194
195 /* Supported request types */
196 enum {
197 NBD_CMD_READ = 0,
198 NBD_CMD_WRITE = 1,
199 NBD_CMD_DISC = 2,
200 NBD_CMD_FLUSH = 3,
201 NBD_CMD_TRIM = 4,
202 NBD_CMD_CACHE = 5,
203 NBD_CMD_WRITE_ZEROES = 6,
204 NBD_CMD_BLOCK_STATUS = 7,
205 };
206
207 #define NBD_DEFAULT_PORT 10809
208
209 /* Maximum size of a single READ/WRITE data buffer */
210 #define NBD_MAX_BUFFER_SIZE (32 * 1024 * 1024)
211
212 /* Maximum size of an export name. The NBD spec requires 256 and
213 * suggests that servers support up to 4096, but we stick to only the
214 * required size so that we can stack-allocate the names, and because
215 * going larger would require an audit of more code to make sure we
216 * aren't overflowing some other buffer. */
217 #define NBD_MAX_NAME_SIZE 256
218
219 /* Two types of reply structures */
220 #define NBD_SIMPLE_REPLY_MAGIC 0x67446698
221 #define NBD_STRUCTURED_REPLY_MAGIC 0x668e33ef
222
223 /* Structured reply flags */
224 #define NBD_REPLY_FLAG_DONE (1 << 0) /* This reply-chunk is last */
225
226 /* Structured reply types */
227 #define NBD_REPLY_ERR(value) ((1 << 15) | (value))
228
229 #define NBD_REPLY_TYPE_NONE 0
230 #define NBD_REPLY_TYPE_OFFSET_DATA 1
231 #define NBD_REPLY_TYPE_OFFSET_HOLE 2
232 #define NBD_REPLY_TYPE_BLOCK_STATUS 5
233 #define NBD_REPLY_TYPE_ERROR NBD_REPLY_ERR(1)
234 #define NBD_REPLY_TYPE_ERROR_OFFSET NBD_REPLY_ERR(2)
235
236 /* Extent flags for base:allocation in NBD_REPLY_TYPE_BLOCK_STATUS */
237 #define NBD_STATE_HOLE (1 << 0)
238 #define NBD_STATE_ZERO (1 << 1)
239
240 /* Extent flags for qemu:dirty-bitmap in NBD_REPLY_TYPE_BLOCK_STATUS */
241 #define NBD_STATE_DIRTY (1 << 0)
242
243 static inline bool nbd_reply_type_is_error(int type)
244 {
245 return type & (1 << 15);
246 }
247
248 /* NBD errors are based on errno numbers, so there is a 1:1 mapping,
249 * but only a limited set of errno values is specified in the protocol.
250 * Everything else is squashed to EINVAL.
251 */
252 #define NBD_SUCCESS 0
253 #define NBD_EPERM 1
254 #define NBD_EIO 5
255 #define NBD_ENOMEM 12
256 #define NBD_EINVAL 22
257 #define NBD_ENOSPC 28
258 #define NBD_EOVERFLOW 75
259 #define NBD_ESHUTDOWN 108
260
261 /* Details collected by NBD_OPT_EXPORT_NAME and NBD_OPT_GO */
262 struct NBDExportInfo {
263 /* Set by client before nbd_receive_negotiate() */
264 bool request_sizes;
265 char *x_dirty_bitmap;
266
267 /* Set by client before nbd_receive_negotiate(), or by server results
268 * during nbd_receive_export_list() */
269 char *name; /* must be non-NULL */
270
271 /* In-out fields, set by client before nbd_receive_negotiate() and
272 * updated by server results during nbd_receive_negotiate() */
273 bool structured_reply;
274 bool base_allocation; /* base:allocation context for NBD_CMD_BLOCK_STATUS */
275
276 /* Set by server results during nbd_receive_negotiate() and
277 * nbd_receive_export_list() */
278 uint64_t size;
279 uint16_t flags;
280 uint32_t min_block;
281 uint32_t opt_block;
282 uint32_t max_block;
283
284 uint32_t context_id;
285
286 /* Set by server results during nbd_receive_export_list() */
287 char *description;
288 int n_contexts;
289 char **contexts;
290 };
291 typedef struct NBDExportInfo NBDExportInfo;
292
293 int nbd_receive_negotiate(QIOChannel *ioc, QCryptoTLSCreds *tlscreds,
294 const char *hostname, QIOChannel **outioc,
295 NBDExportInfo *info, Error **errp);
296 void nbd_free_export_list(NBDExportInfo *info, int count);
297 int nbd_receive_export_list(QIOChannel *ioc, QCryptoTLSCreds *tlscreds,
298 const char *hostname, NBDExportInfo **info,
299 Error **errp);
300 int nbd_init(int fd, QIOChannelSocket *sioc, NBDExportInfo *info,
301 Error **errp);
302 int nbd_send_request(QIOChannel *ioc, NBDRequest *request);
303 int coroutine_fn nbd_receive_reply(BlockDriverState *bs, QIOChannel *ioc,
304 NBDReply *reply, Error **errp);
305 int nbd_client(int fd);
306 int nbd_disconnect(int fd);
307 int nbd_errno_to_system_errno(int err);
308
309 typedef struct NBDExport NBDExport;
310 typedef struct NBDClient NBDClient;
311
312 NBDExport *nbd_export_new(BlockDriverState *bs, uint64_t dev_offset,
313 uint64_t size, const char *name, const char *desc,
314 const char *bitmap, uint16_t nbdflags,
315 void (*close)(NBDExport *), bool writethrough,
316 BlockBackend *on_eject_blk, Error **errp);
317 void nbd_export_close(NBDExport *exp);
318 void nbd_export_remove(NBDExport *exp, NbdServerRemoveMode mode, Error **errp);
319 void nbd_export_get(NBDExport *exp);
320 void nbd_export_put(NBDExport *exp);
321
322 BlockBackend *nbd_export_get_blockdev(NBDExport *exp);
323
324 NBDExport *nbd_export_find(const char *name);
325 void nbd_export_close_all(void);
326
327 void nbd_client_new(QIOChannelSocket *sioc,
328 QCryptoTLSCreds *tlscreds,
329 const char *tlsauthz,
330 void (*close_fn)(NBDClient *, bool));
331 void nbd_client_get(NBDClient *client);
332 void nbd_client_put(NBDClient *client);
333
334 void nbd_server_start(SocketAddress *addr, const char *tls_creds,
335 const char *tls_authz, Error **errp);
336
337 /* nbd_read
338 * Reads @size bytes from @ioc. Returns 0 on success.
339 */
340 static inline int nbd_read(QIOChannel *ioc, void *buffer, size_t size,
341 const char *desc, Error **errp)
342 {
343 int ret = qio_channel_read_all(ioc, buffer, size, errp) < 0 ? -EIO : 0;
344
345 if (ret < 0) {
346 if (desc) {
347 error_prepend(errp, "Failed to read %s: ", desc);
348 }
349 return -1;
350 }
351
352 return 0;
353 }
354
355 #define DEF_NBD_READ_N(bits) \
356 static inline int nbd_read##bits(QIOChannel *ioc, \
357 uint##bits##_t *val, \
358 const char *desc, Error **errp) \
359 { \
360 if (nbd_read(ioc, val, sizeof(*val), desc, errp) < 0) { \
361 return -1; \
362 } \
363 *val = be##bits##_to_cpu(*val); \
364 return 0; \
365 }
366
367 DEF_NBD_READ_N(16) /* Defines nbd_read16(). */
368 DEF_NBD_READ_N(32) /* Defines nbd_read32(). */
369 DEF_NBD_READ_N(64) /* Defines nbd_read64(). */
370
371 #undef DEF_NBD_READ_N
372
373 static inline bool nbd_reply_is_simple(NBDReply *reply)
374 {
375 return reply->magic == NBD_SIMPLE_REPLY_MAGIC;
376 }
377
378 static inline bool nbd_reply_is_structured(NBDReply *reply)
379 {
380 return reply->magic == NBD_STRUCTURED_REPLY_MAGIC;
381 }
382
383 const char *nbd_reply_type_lookup(uint16_t type);
384 const char *nbd_opt_lookup(uint32_t opt);
385 const char *nbd_rep_lookup(uint32_t rep);
386 const char *nbd_info_lookup(uint16_t info);
387 const char *nbd_cmd_lookup(uint16_t info);
388 const char *nbd_err_lookup(int err);
389
390 #endif