Merge remote-tracking branch 'remotes/jasowang/tags/net-pull-request' into staging
[qemu.git] / include / block / nbd.h
1 /*
2 * Copyright (C) 2016-2020 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 "block/export.h"
24 #include "io/channel-socket.h"
25 #include "crypto/tlscreds.h"
26 #include "qapi/error.h"
27
28 extern const BlockExportDriver blk_exp_nbd;
29
30 /* Handshake phase structs - this struct is passed on the wire */
31
32 struct NBDOption {
33 uint64_t magic; /* NBD_OPTS_MAGIC */
34 uint32_t option; /* NBD_OPT_* */
35 uint32_t length;
36 } QEMU_PACKED;
37 typedef struct NBDOption NBDOption;
38
39 struct NBDOptionReply {
40 uint64_t magic; /* NBD_REP_MAGIC */
41 uint32_t option; /* NBD_OPT_* */
42 uint32_t type; /* NBD_REP_* */
43 uint32_t length;
44 } QEMU_PACKED;
45 typedef struct NBDOptionReply NBDOptionReply;
46
47 typedef struct NBDOptionReplyMetaContext {
48 NBDOptionReply h; /* h.type = NBD_REP_META_CONTEXT, h.length > 4 */
49 uint32_t context_id;
50 /* metadata context name follows */
51 } QEMU_PACKED NBDOptionReplyMetaContext;
52
53 /* Transmission phase structs
54 *
55 * Note: these are _NOT_ the same as the network representation of an NBD
56 * request and reply!
57 */
58 struct NBDRequest {
59 uint64_t handle;
60 uint64_t from;
61 uint32_t len;
62 uint16_t flags; /* NBD_CMD_FLAG_* */
63 uint16_t type; /* NBD_CMD_* */
64 };
65 typedef struct NBDRequest NBDRequest;
66
67 typedef struct NBDSimpleReply {
68 uint32_t magic; /* NBD_SIMPLE_REPLY_MAGIC */
69 uint32_t error;
70 uint64_t handle;
71 } QEMU_PACKED NBDSimpleReply;
72
73 /* Header of all structured replies */
74 typedef struct NBDStructuredReplyChunk {
75 uint32_t magic; /* NBD_STRUCTURED_REPLY_MAGIC */
76 uint16_t flags; /* combination of NBD_REPLY_FLAG_* */
77 uint16_t type; /* NBD_REPLY_TYPE_* */
78 uint64_t handle; /* request handle */
79 uint32_t length; /* length of payload */
80 } QEMU_PACKED NBDStructuredReplyChunk;
81
82 typedef union NBDReply {
83 NBDSimpleReply simple;
84 NBDStructuredReplyChunk structured;
85 struct {
86 /* @magic and @handle fields have the same offset and size both in
87 * simple reply and structured reply chunk, so let them be accessible
88 * without ".simple." or ".structured." specification
89 */
90 uint32_t magic;
91 uint32_t _skip;
92 uint64_t handle;
93 } QEMU_PACKED;
94 } NBDReply;
95
96 /* Header of chunk for NBD_REPLY_TYPE_OFFSET_DATA */
97 typedef struct NBDStructuredReadData {
98 NBDStructuredReplyChunk h; /* h.length >= 9 */
99 uint64_t offset;
100 /* At least one byte of data payload follows, calculated from h.length */
101 } QEMU_PACKED NBDStructuredReadData;
102
103 /* Complete chunk for NBD_REPLY_TYPE_OFFSET_HOLE */
104 typedef struct NBDStructuredReadHole {
105 NBDStructuredReplyChunk h; /* h.length == 12 */
106 uint64_t offset;
107 uint32_t length;
108 } QEMU_PACKED NBDStructuredReadHole;
109
110 /* Header of all NBD_REPLY_TYPE_ERROR* errors */
111 typedef struct NBDStructuredError {
112 NBDStructuredReplyChunk h; /* h.length >= 6 */
113 uint32_t error;
114 uint16_t message_length;
115 } QEMU_PACKED NBDStructuredError;
116
117 /* Header of NBD_REPLY_TYPE_BLOCK_STATUS */
118 typedef struct NBDStructuredMeta {
119 NBDStructuredReplyChunk h; /* h.length >= 12 (at least one extent) */
120 uint32_t context_id;
121 /* extents follows */
122 } QEMU_PACKED NBDStructuredMeta;
123
124 /* Extent chunk for NBD_REPLY_TYPE_BLOCK_STATUS */
125 typedef struct NBDExtent {
126 uint32_t length;
127 uint32_t flags; /* NBD_STATE_* */
128 } QEMU_PACKED NBDExtent;
129
130 /* Transmission (export) flags: sent from server to client during handshake,
131 but describe what will happen during transmission */
132 enum {
133 NBD_FLAG_HAS_FLAGS_BIT = 0, /* Flags are there */
134 NBD_FLAG_READ_ONLY_BIT = 1, /* Device is read-only */
135 NBD_FLAG_SEND_FLUSH_BIT = 2, /* Send FLUSH */
136 NBD_FLAG_SEND_FUA_BIT = 3, /* Send FUA (Force Unit Access) */
137 NBD_FLAG_ROTATIONAL_BIT = 4, /* Use elevator algorithm -
138 rotational media */
139 NBD_FLAG_SEND_TRIM_BIT = 5, /* Send TRIM (discard) */
140 NBD_FLAG_SEND_WRITE_ZEROES_BIT = 6, /* Send WRITE_ZEROES */
141 NBD_FLAG_SEND_DF_BIT = 7, /* Send DF (Do not Fragment) */
142 NBD_FLAG_CAN_MULTI_CONN_BIT = 8, /* Multi-client cache consistent */
143 NBD_FLAG_SEND_RESIZE_BIT = 9, /* Send resize */
144 NBD_FLAG_SEND_CACHE_BIT = 10, /* Send CACHE (prefetch) */
145 NBD_FLAG_SEND_FAST_ZERO_BIT = 11, /* FAST_ZERO flag for WRITE_ZEROES */
146 };
147
148 #define NBD_FLAG_HAS_FLAGS (1 << NBD_FLAG_HAS_FLAGS_BIT)
149 #define NBD_FLAG_READ_ONLY (1 << NBD_FLAG_READ_ONLY_BIT)
150 #define NBD_FLAG_SEND_FLUSH (1 << NBD_FLAG_SEND_FLUSH_BIT)
151 #define NBD_FLAG_SEND_FUA (1 << NBD_FLAG_SEND_FUA_BIT)
152 #define NBD_FLAG_ROTATIONAL (1 << NBD_FLAG_ROTATIONAL_BIT)
153 #define NBD_FLAG_SEND_TRIM (1 << NBD_FLAG_SEND_TRIM_BIT)
154 #define NBD_FLAG_SEND_WRITE_ZEROES (1 << NBD_FLAG_SEND_WRITE_ZEROES_BIT)
155 #define NBD_FLAG_SEND_DF (1 << NBD_FLAG_SEND_DF_BIT)
156 #define NBD_FLAG_CAN_MULTI_CONN (1 << NBD_FLAG_CAN_MULTI_CONN_BIT)
157 #define NBD_FLAG_SEND_RESIZE (1 << NBD_FLAG_SEND_RESIZE_BIT)
158 #define NBD_FLAG_SEND_CACHE (1 << NBD_FLAG_SEND_CACHE_BIT)
159 #define NBD_FLAG_SEND_FAST_ZERO (1 << NBD_FLAG_SEND_FAST_ZERO_BIT)
160
161 /* New-style handshake (global) flags, sent from server to client, and
162 control what will happen during handshake phase. */
163 #define NBD_FLAG_FIXED_NEWSTYLE (1 << 0) /* Fixed newstyle protocol. */
164 #define NBD_FLAG_NO_ZEROES (1 << 1) /* End handshake without zeroes. */
165
166 /* New-style client flags, sent from client to server to control what happens
167 during handshake phase. */
168 #define NBD_FLAG_C_FIXED_NEWSTYLE (1 << 0) /* Fixed newstyle protocol. */
169 #define NBD_FLAG_C_NO_ZEROES (1 << 1) /* End handshake without zeroes. */
170
171 /* Option requests. */
172 #define NBD_OPT_EXPORT_NAME (1)
173 #define NBD_OPT_ABORT (2)
174 #define NBD_OPT_LIST (3)
175 /* #define NBD_OPT_PEEK_EXPORT (4) not in use */
176 #define NBD_OPT_STARTTLS (5)
177 #define NBD_OPT_INFO (6)
178 #define NBD_OPT_GO (7)
179 #define NBD_OPT_STRUCTURED_REPLY (8)
180 #define NBD_OPT_LIST_META_CONTEXT (9)
181 #define NBD_OPT_SET_META_CONTEXT (10)
182
183 /* Option reply types. */
184 #define NBD_REP_ERR(value) ((UINT32_C(1) << 31) | (value))
185
186 #define NBD_REP_ACK (1) /* Data sending finished. */
187 #define NBD_REP_SERVER (2) /* Export description. */
188 #define NBD_REP_INFO (3) /* NBD_OPT_INFO/GO. */
189 #define NBD_REP_META_CONTEXT (4) /* NBD_OPT_{LIST,SET}_META_CONTEXT */
190
191 #define NBD_REP_ERR_UNSUP NBD_REP_ERR(1) /* Unknown option */
192 #define NBD_REP_ERR_POLICY NBD_REP_ERR(2) /* Server denied */
193 #define NBD_REP_ERR_INVALID NBD_REP_ERR(3) /* Invalid length */
194 #define NBD_REP_ERR_PLATFORM NBD_REP_ERR(4) /* Not compiled in */
195 #define NBD_REP_ERR_TLS_REQD NBD_REP_ERR(5) /* TLS required */
196 #define NBD_REP_ERR_UNKNOWN NBD_REP_ERR(6) /* Export unknown */
197 #define NBD_REP_ERR_SHUTDOWN NBD_REP_ERR(7) /* Server shutting down */
198 #define NBD_REP_ERR_BLOCK_SIZE_REQD NBD_REP_ERR(8) /* Need INFO_BLOCK_SIZE */
199
200 /* Info types, used during NBD_REP_INFO */
201 #define NBD_INFO_EXPORT 0
202 #define NBD_INFO_NAME 1
203 #define NBD_INFO_DESCRIPTION 2
204 #define NBD_INFO_BLOCK_SIZE 3
205
206 /* Request flags, sent from client to server during transmission phase */
207 #define NBD_CMD_FLAG_FUA (1 << 0) /* 'force unit access' during write */
208 #define NBD_CMD_FLAG_NO_HOLE (1 << 1) /* don't punch hole on zero run */
209 #define NBD_CMD_FLAG_DF (1 << 2) /* don't fragment structured read */
210 #define NBD_CMD_FLAG_REQ_ONE (1 << 3) /* only one extent in BLOCK_STATUS
211 * reply chunk */
212 #define NBD_CMD_FLAG_FAST_ZERO (1 << 4) /* fail if WRITE_ZEROES is not fast */
213
214 /* Supported request types */
215 enum {
216 NBD_CMD_READ = 0,
217 NBD_CMD_WRITE = 1,
218 NBD_CMD_DISC = 2,
219 NBD_CMD_FLUSH = 3,
220 NBD_CMD_TRIM = 4,
221 NBD_CMD_CACHE = 5,
222 NBD_CMD_WRITE_ZEROES = 6,
223 NBD_CMD_BLOCK_STATUS = 7,
224 };
225
226 #define NBD_DEFAULT_PORT 10809
227
228 /* Maximum size of a single READ/WRITE data buffer */
229 #define NBD_MAX_BUFFER_SIZE (32 * 1024 * 1024)
230
231 /*
232 * Maximum size of a protocol string (export name, metadata context name,
233 * etc.). Use malloc rather than stack allocation for storage of a
234 * string.
235 */
236 #define NBD_MAX_STRING_SIZE 4096
237
238 /* Two types of reply structures */
239 #define NBD_SIMPLE_REPLY_MAGIC 0x67446698
240 #define NBD_STRUCTURED_REPLY_MAGIC 0x668e33ef
241
242 /* Structured reply flags */
243 #define NBD_REPLY_FLAG_DONE (1 << 0) /* This reply-chunk is last */
244
245 /* Structured reply types */
246 #define NBD_REPLY_ERR(value) ((1 << 15) | (value))
247
248 #define NBD_REPLY_TYPE_NONE 0
249 #define NBD_REPLY_TYPE_OFFSET_DATA 1
250 #define NBD_REPLY_TYPE_OFFSET_HOLE 2
251 #define NBD_REPLY_TYPE_BLOCK_STATUS 5
252 #define NBD_REPLY_TYPE_ERROR NBD_REPLY_ERR(1)
253 #define NBD_REPLY_TYPE_ERROR_OFFSET NBD_REPLY_ERR(2)
254
255 /* Extent flags for base:allocation in NBD_REPLY_TYPE_BLOCK_STATUS */
256 #define NBD_STATE_HOLE (1 << 0)
257 #define NBD_STATE_ZERO (1 << 1)
258
259 /* Extent flags for qemu:dirty-bitmap in NBD_REPLY_TYPE_BLOCK_STATUS */
260 #define NBD_STATE_DIRTY (1 << 0)
261
262 /* No flags needed for qemu:allocation-depth in NBD_REPLY_TYPE_BLOCK_STATUS */
263
264 static inline bool nbd_reply_type_is_error(int type)
265 {
266 return type & (1 << 15);
267 }
268
269 /* NBD errors are based on errno numbers, so there is a 1:1 mapping,
270 * but only a limited set of errno values is specified in the protocol.
271 * Everything else is squashed to EINVAL.
272 */
273 #define NBD_SUCCESS 0
274 #define NBD_EPERM 1
275 #define NBD_EIO 5
276 #define NBD_ENOMEM 12
277 #define NBD_EINVAL 22
278 #define NBD_ENOSPC 28
279 #define NBD_EOVERFLOW 75
280 #define NBD_ENOTSUP 95
281 #define NBD_ESHUTDOWN 108
282
283 /* Details collected by NBD_OPT_EXPORT_NAME and NBD_OPT_GO */
284 struct NBDExportInfo {
285 /* Set by client before nbd_receive_negotiate() */
286 bool request_sizes;
287 char *x_dirty_bitmap;
288
289 /* Set by client before nbd_receive_negotiate(), or by server results
290 * during nbd_receive_export_list() */
291 char *name; /* must be non-NULL */
292
293 /* In-out fields, set by client before nbd_receive_negotiate() and
294 * updated by server results during nbd_receive_negotiate() */
295 bool structured_reply;
296 bool base_allocation; /* base:allocation context for NBD_CMD_BLOCK_STATUS */
297
298 /* Set by server results during nbd_receive_negotiate() and
299 * nbd_receive_export_list() */
300 uint64_t size;
301 uint16_t flags;
302 uint32_t min_block;
303 uint32_t opt_block;
304 uint32_t max_block;
305
306 uint32_t context_id;
307
308 /* Set by server results during nbd_receive_export_list() */
309 char *description;
310 int n_contexts;
311 char **contexts;
312 };
313 typedef struct NBDExportInfo NBDExportInfo;
314
315 int nbd_receive_negotiate(AioContext *aio_context, QIOChannel *ioc,
316 QCryptoTLSCreds *tlscreds,
317 const char *hostname, QIOChannel **outioc,
318 NBDExportInfo *info, Error **errp);
319 void nbd_free_export_list(NBDExportInfo *info, int count);
320 int nbd_receive_export_list(QIOChannel *ioc, QCryptoTLSCreds *tlscreds,
321 const char *hostname, NBDExportInfo **info,
322 Error **errp);
323 int nbd_init(int fd, QIOChannelSocket *sioc, NBDExportInfo *info,
324 Error **errp);
325 int nbd_send_request(QIOChannel *ioc, NBDRequest *request);
326 int coroutine_fn nbd_receive_reply(BlockDriverState *bs, QIOChannel *ioc,
327 NBDReply *reply, Error **errp);
328 int nbd_client(int fd);
329 int nbd_disconnect(int fd);
330 int nbd_errno_to_system_errno(int err);
331
332 typedef struct NBDExport NBDExport;
333 typedef struct NBDClient NBDClient;
334
335 void nbd_export_set_on_eject_blk(BlockExport *exp, BlockBackend *blk);
336
337 AioContext *nbd_export_aio_context(NBDExport *exp);
338 NBDExport *nbd_export_find(const char *name);
339
340 void nbd_client_new(QIOChannelSocket *sioc,
341 QCryptoTLSCreds *tlscreds,
342 const char *tlsauthz,
343 void (*close_fn)(NBDClient *, bool));
344 void nbd_client_get(NBDClient *client);
345 void nbd_client_put(NBDClient *client);
346
347 void nbd_server_is_qemu_nbd(bool value);
348 bool nbd_server_is_running(void);
349 void nbd_server_start(SocketAddress *addr, const char *tls_creds,
350 const char *tls_authz, uint32_t max_connections,
351 Error **errp);
352 void nbd_server_start_options(NbdServerOptions *arg, Error **errp);
353
354 /* nbd_read
355 * Reads @size bytes from @ioc. Returns 0 on success.
356 */
357 static inline int nbd_read(QIOChannel *ioc, void *buffer, size_t size,
358 const char *desc, Error **errp)
359 {
360 ERRP_GUARD();
361 int ret = qio_channel_read_all(ioc, buffer, size, errp) < 0 ? -EIO : 0;
362
363 if (ret < 0) {
364 if (desc) {
365 error_prepend(errp, "Failed to read %s: ", desc);
366 }
367 return ret;
368 }
369
370 return 0;
371 }
372
373 #define DEF_NBD_READ_N(bits) \
374 static inline int nbd_read##bits(QIOChannel *ioc, \
375 uint##bits##_t *val, \
376 const char *desc, Error **errp) \
377 { \
378 int ret = nbd_read(ioc, val, sizeof(*val), desc, errp); \
379 if (ret < 0) { \
380 return ret; \
381 } \
382 *val = be##bits##_to_cpu(*val); \
383 return 0; \
384 }
385
386 DEF_NBD_READ_N(16) /* Defines nbd_read16(). */
387 DEF_NBD_READ_N(32) /* Defines nbd_read32(). */
388 DEF_NBD_READ_N(64) /* Defines nbd_read64(). */
389
390 #undef DEF_NBD_READ_N
391
392 static inline bool nbd_reply_is_simple(NBDReply *reply)
393 {
394 return reply->magic == NBD_SIMPLE_REPLY_MAGIC;
395 }
396
397 static inline bool nbd_reply_is_structured(NBDReply *reply)
398 {
399 return reply->magic == NBD_STRUCTURED_REPLY_MAGIC;
400 }
401
402 const char *nbd_reply_type_lookup(uint16_t type);
403 const char *nbd_opt_lookup(uint32_t opt);
404 const char *nbd_rep_lookup(uint32_t rep);
405 const char *nbd_info_lookup(uint16_t info);
406 const char *nbd_cmd_lookup(uint16_t info);
407 const char *nbd_err_lookup(int err);
408
409 #endif