1 /* 2 * Copyright (C) 2016-2022 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 #include "qemu/bswap.h" 28 29 extern const BlockExportDriver blk_exp_nbd; 30 31 /* Handshake phase structs - this struct is passed on the wire */ 32 33 struct NBDOption { 34 uint64_t magic; /* NBD_OPTS_MAGIC */ 35 uint32_t option; /* NBD_OPT_* */ 36 uint32_t length; 37 } QEMU_PACKED; 38 typedef struct NBDOption NBDOption; 39 40 struct NBDOptionReply { 41 uint64_t magic; /* NBD_REP_MAGIC */ 42 uint32_t option; /* NBD_OPT_* */ 43 uint32_t type; /* NBD_REP_* */ 44 uint32_t length; 45 } QEMU_PACKED; 46 typedef struct NBDOptionReply NBDOptionReply; 47 48 typedef struct NBDOptionReplyMetaContext { 49 NBDOptionReply h; /* h.type = NBD_REP_META_CONTEXT, h.length > 4 */ 50 uint32_t context_id; 51 /* metadata context name follows */ 52 } QEMU_PACKED NBDOptionReplyMetaContext; 53 54 /* Transmission phase structs 55 * 56 * Note: these are _NOT_ the same as the network representation of an NBD 57 * request and reply! 58 */ 59 struct NBDRequest { 60 uint64_t handle; 61 uint64_t from; 62 uint32_t len; 63 uint16_t flags; /* NBD_CMD_FLAG_* */ 64 uint16_t type; /* NBD_CMD_* */ 65 }; 66 typedef struct NBDRequest NBDRequest; 67 68 typedef struct NBDSimpleReply { 69 uint32_t magic; /* NBD_SIMPLE_REPLY_MAGIC */ 70 uint32_t error; 71 uint64_t handle; 72 } QEMU_PACKED NBDSimpleReply; 73 74 /* Header of all structured replies */ 75 typedef struct NBDStructuredReplyChunk { 76 uint32_t magic; /* NBD_STRUCTURED_REPLY_MAGIC */ 77 uint16_t flags; /* combination of NBD_REPLY_FLAG_* */ 78 uint16_t type; /* NBD_REPLY_TYPE_* */ 79 uint64_t handle; /* request handle */ 80 uint32_t length; /* length of payload */ 81 } QEMU_PACKED NBDStructuredReplyChunk; 82 83 typedef union NBDReply { 84 NBDSimpleReply simple; 85 NBDStructuredReplyChunk structured; 86 struct { 87 /* @magic and @handle fields have the same offset and size both in 88 * simple reply and structured reply chunk, so let them be accessible 89 * without ".simple." or ".structured." specification 90 */ 91 uint32_t magic; 92 uint32_t _skip; 93 uint64_t handle; 94 } QEMU_PACKED; 95 } NBDReply; 96 97 /* Header of chunk for NBD_REPLY_TYPE_OFFSET_DATA */ 98 typedef struct NBDStructuredReadData { 99 NBDStructuredReplyChunk h; /* h.length >= 9 */ 100 uint64_t offset; 101 /* At least one byte of data payload follows, calculated from h.length */ 102 } QEMU_PACKED NBDStructuredReadData; 103 104 /* Complete chunk for NBD_REPLY_TYPE_OFFSET_HOLE */ 105 typedef struct NBDStructuredReadHole { 106 NBDStructuredReplyChunk h; /* h.length == 12 */ 107 uint64_t offset; 108 uint32_t length; 109 } QEMU_PACKED NBDStructuredReadHole; 110 111 /* Header of all NBD_REPLY_TYPE_ERROR* errors */ 112 typedef struct NBDStructuredError { 113 NBDStructuredReplyChunk h; /* h.length >= 6 */ 114 uint32_t error; 115 uint16_t message_length; 116 } QEMU_PACKED NBDStructuredError; 117 118 /* Header of NBD_REPLY_TYPE_BLOCK_STATUS */ 119 typedef struct NBDStructuredMeta { 120 NBDStructuredReplyChunk h; /* h.length >= 12 (at least one extent) */ 121 uint32_t context_id; 122 /* extents follows */ 123 } QEMU_PACKED NBDStructuredMeta; 124 125 /* Extent chunk for NBD_REPLY_TYPE_BLOCK_STATUS */ 126 typedef struct NBDExtent { 127 uint32_t length; 128 uint32_t flags; /* NBD_STATE_* */ 129 } QEMU_PACKED NBDExtent; 130 131 /* Transmission (export) flags: sent from server to client during handshake, 132 but describe what will happen during transmission */ 133 enum { 134 NBD_FLAG_HAS_FLAGS_BIT = 0, /* Flags are there */ 135 NBD_FLAG_READ_ONLY_BIT = 1, /* Device is read-only */ 136 NBD_FLAG_SEND_FLUSH_BIT = 2, /* Send FLUSH */ 137 NBD_FLAG_SEND_FUA_BIT = 3, /* Send FUA (Force Unit Access) */ 138 NBD_FLAG_ROTATIONAL_BIT = 4, /* Use elevator algorithm - 139 rotational media */ 140 NBD_FLAG_SEND_TRIM_BIT = 5, /* Send TRIM (discard) */ 141 NBD_FLAG_SEND_WRITE_ZEROES_BIT = 6, /* Send WRITE_ZEROES */ 142 NBD_FLAG_SEND_DF_BIT = 7, /* Send DF (Do not Fragment) */ 143 NBD_FLAG_CAN_MULTI_CONN_BIT = 8, /* Multi-client cache consistent */ 144 NBD_FLAG_SEND_RESIZE_BIT = 9, /* Send resize */ 145 NBD_FLAG_SEND_CACHE_BIT = 10, /* Send CACHE (prefetch) */ 146 NBD_FLAG_SEND_FAST_ZERO_BIT = 11, /* FAST_ZERO flag for WRITE_ZEROES */ 147 }; 148 149 #define NBD_FLAG_HAS_FLAGS (1 << NBD_FLAG_HAS_FLAGS_BIT) 150 #define NBD_FLAG_READ_ONLY (1 << NBD_FLAG_READ_ONLY_BIT) 151 #define NBD_FLAG_SEND_FLUSH (1 << NBD_FLAG_SEND_FLUSH_BIT) 152 #define NBD_FLAG_SEND_FUA (1 << NBD_FLAG_SEND_FUA_BIT) 153 #define NBD_FLAG_ROTATIONAL (1 << NBD_FLAG_ROTATIONAL_BIT) 154 #define NBD_FLAG_SEND_TRIM (1 << NBD_FLAG_SEND_TRIM_BIT) 155 #define NBD_FLAG_SEND_WRITE_ZEROES (1 << NBD_FLAG_SEND_WRITE_ZEROES_BIT) 156 #define NBD_FLAG_SEND_DF (1 << NBD_FLAG_SEND_DF_BIT) 157 #define NBD_FLAG_CAN_MULTI_CONN (1 << NBD_FLAG_CAN_MULTI_CONN_BIT) 158 #define NBD_FLAG_SEND_RESIZE (1 << NBD_FLAG_SEND_RESIZE_BIT) 159 #define NBD_FLAG_SEND_CACHE (1 << NBD_FLAG_SEND_CACHE_BIT) 160 #define NBD_FLAG_SEND_FAST_ZERO (1 << NBD_FLAG_SEND_FAST_ZERO_BIT) 161 162 /* New-style handshake (global) flags, sent from server to client, and 163 control what will happen during handshake phase. */ 164 #define NBD_FLAG_FIXED_NEWSTYLE (1 << 0) /* Fixed newstyle protocol. */ 165 #define NBD_FLAG_NO_ZEROES (1 << 1) /* End handshake without zeroes. */ 166 167 /* New-style client flags, sent from client to server to control what happens 168 during handshake phase. */ 169 #define NBD_FLAG_C_FIXED_NEWSTYLE (1 << 0) /* Fixed newstyle protocol. */ 170 #define NBD_FLAG_C_NO_ZEROES (1 << 1) /* End handshake without zeroes. */ 171 172 /* Option requests. */ 173 #define NBD_OPT_EXPORT_NAME (1) 174 #define NBD_OPT_ABORT (2) 175 #define NBD_OPT_LIST (3) 176 /* #define NBD_OPT_PEEK_EXPORT (4) not in use */ 177 #define NBD_OPT_STARTTLS (5) 178 #define NBD_OPT_INFO (6) 179 #define NBD_OPT_GO (7) 180 #define NBD_OPT_STRUCTURED_REPLY (8) 181 #define NBD_OPT_LIST_META_CONTEXT (9) 182 #define NBD_OPT_SET_META_CONTEXT (10) 183 184 /* Option reply types. */ 185 #define NBD_REP_ERR(value) ((UINT32_C(1) << 31) | (value)) 186 187 #define NBD_REP_ACK (1) /* Data sending finished. */ 188 #define NBD_REP_SERVER (2) /* Export description. */ 189 #define NBD_REP_INFO (3) /* NBD_OPT_INFO/GO. */ 190 #define NBD_REP_META_CONTEXT (4) /* NBD_OPT_{LIST,SET}_META_CONTEXT */ 191 192 #define NBD_REP_ERR_UNSUP NBD_REP_ERR(1) /* Unknown option */ 193 #define NBD_REP_ERR_POLICY NBD_REP_ERR(2) /* Server denied */ 194 #define NBD_REP_ERR_INVALID NBD_REP_ERR(3) /* Invalid length */ 195 #define NBD_REP_ERR_PLATFORM NBD_REP_ERR(4) /* Not compiled in */ 196 #define NBD_REP_ERR_TLS_REQD NBD_REP_ERR(5) /* TLS required */ 197 #define NBD_REP_ERR_UNKNOWN NBD_REP_ERR(6) /* Export unknown */ 198 #define NBD_REP_ERR_SHUTDOWN NBD_REP_ERR(7) /* Server shutting down */ 199 #define NBD_REP_ERR_BLOCK_SIZE_REQD NBD_REP_ERR(8) /* Need INFO_BLOCK_SIZE */ 200 201 /* Info types, used during NBD_REP_INFO */ 202 #define NBD_INFO_EXPORT 0 203 #define NBD_INFO_NAME 1 204 #define NBD_INFO_DESCRIPTION 2 205 #define NBD_INFO_BLOCK_SIZE 3 206 207 /* Request flags, sent from client to server during transmission phase */ 208 #define NBD_CMD_FLAG_FUA (1 << 0) /* 'force unit access' during write */ 209 #define NBD_CMD_FLAG_NO_HOLE (1 << 1) /* don't punch hole on zero run */ 210 #define NBD_CMD_FLAG_DF (1 << 2) /* don't fragment structured read */ 211 #define NBD_CMD_FLAG_REQ_ONE (1 << 3) /* only one extent in BLOCK_STATUS 212 * reply chunk */ 213 #define NBD_CMD_FLAG_FAST_ZERO (1 << 4) /* fail if WRITE_ZEROES is not fast */ 214 215 /* Supported request types */ 216 enum { 217 NBD_CMD_READ = 0, 218 NBD_CMD_WRITE = 1, 219 NBD_CMD_DISC = 2, 220 NBD_CMD_FLUSH = 3, 221 NBD_CMD_TRIM = 4, 222 NBD_CMD_CACHE = 5, 223 NBD_CMD_WRITE_ZEROES = 6, 224 NBD_CMD_BLOCK_STATUS = 7, 225 }; 226 227 #define NBD_DEFAULT_PORT 10809 228 229 /* Maximum size of a single READ/WRITE data buffer */ 230 #define NBD_MAX_BUFFER_SIZE (32 * 1024 * 1024) 231 232 /* 233 * Maximum size of a protocol string (export name, metadata context name, 234 * etc.). Use malloc rather than stack allocation for storage of a 235 * string. 236 */ 237 #define NBD_MAX_STRING_SIZE 4096 238 239 /* Two types of reply structures */ 240 #define NBD_SIMPLE_REPLY_MAGIC 0x67446698 241 #define NBD_STRUCTURED_REPLY_MAGIC 0x668e33ef 242 243 /* Structured reply flags */ 244 #define NBD_REPLY_FLAG_DONE (1 << 0) /* This reply-chunk is last */ 245 246 /* Structured reply types */ 247 #define NBD_REPLY_ERR(value) ((1 << 15) | (value)) 248 249 #define NBD_REPLY_TYPE_NONE 0 250 #define NBD_REPLY_TYPE_OFFSET_DATA 1 251 #define NBD_REPLY_TYPE_OFFSET_HOLE 2 252 #define NBD_REPLY_TYPE_BLOCK_STATUS 5 253 #define NBD_REPLY_TYPE_ERROR NBD_REPLY_ERR(1) 254 #define NBD_REPLY_TYPE_ERROR_OFFSET NBD_REPLY_ERR(2) 255 256 /* Extent flags for base:allocation in NBD_REPLY_TYPE_BLOCK_STATUS */ 257 #define NBD_STATE_HOLE (1 << 0) 258 #define NBD_STATE_ZERO (1 << 1) 259 260 /* Extent flags for qemu:dirty-bitmap in NBD_REPLY_TYPE_BLOCK_STATUS */ 261 #define NBD_STATE_DIRTY (1 << 0) 262 263 /* No flags needed for qemu:allocation-depth in NBD_REPLY_TYPE_BLOCK_STATUS */ 264 265 static inline bool nbd_reply_type_is_error(int type) 266 { 267 return type & (1 << 15); 268 } 269 270 /* NBD errors are based on errno numbers, so there is a 1:1 mapping, 271 * but only a limited set of errno values is specified in the protocol. 272 * Everything else is squashed to EINVAL. 273 */ 274 #define NBD_SUCCESS 0 275 #define NBD_EPERM 1 276 #define NBD_EIO 5 277 #define NBD_ENOMEM 12 278 #define NBD_EINVAL 22 279 #define NBD_ENOSPC 28 280 #define NBD_EOVERFLOW 75 281 #define NBD_ENOTSUP 95 282 #define NBD_ESHUTDOWN 108 283 284 /* Details collected by NBD_OPT_EXPORT_NAME and NBD_OPT_GO */ 285 struct NBDExportInfo { 286 /* Set by client before nbd_receive_negotiate() */ 287 bool request_sizes; 288 char *x_dirty_bitmap; 289 290 /* Set by client before nbd_receive_negotiate(), or by server results 291 * during nbd_receive_export_list() */ 292 char *name; /* must be non-NULL */ 293 294 /* In-out fields, set by client before nbd_receive_negotiate() and 295 * updated by server results during nbd_receive_negotiate() */ 296 bool structured_reply; 297 bool base_allocation; /* base:allocation context for NBD_CMD_BLOCK_STATUS */ 298 299 /* Set by server results during nbd_receive_negotiate() and 300 * nbd_receive_export_list() */ 301 uint64_t size; 302 uint16_t flags; 303 uint32_t min_block; 304 uint32_t opt_block; 305 uint32_t max_block; 306 307 uint32_t context_id; 308 309 /* Set by server results during nbd_receive_export_list() */ 310 char *description; 311 int n_contexts; 312 char **contexts; 313 }; 314 typedef struct NBDExportInfo NBDExportInfo; 315 316 int nbd_receive_negotiate(AioContext *aio_context, QIOChannel *ioc, 317 QCryptoTLSCreds *tlscreds, 318 const char *hostname, QIOChannel **outioc, 319 NBDExportInfo *info, Error **errp); 320 void nbd_free_export_list(NBDExportInfo *info, int count); 321 int nbd_receive_export_list(QIOChannel *ioc, QCryptoTLSCreds *tlscreds, 322 const char *hostname, NBDExportInfo **info, 323 Error **errp); 324 int nbd_init(int fd, QIOChannelSocket *sioc, NBDExportInfo *info, 325 Error **errp); 326 int nbd_send_request(QIOChannel *ioc, NBDRequest *request); 327 int coroutine_fn nbd_receive_reply(BlockDriverState *bs, QIOChannel *ioc, 328 NBDReply *reply, Error **errp); 329 int nbd_client(int fd); 330 int nbd_disconnect(int fd); 331 int nbd_errno_to_system_errno(int err); 332 333 typedef struct NBDExport NBDExport; 334 typedef struct NBDClient NBDClient; 335 336 void nbd_export_set_on_eject_blk(BlockExport *exp, BlockBackend *blk); 337 338 AioContext *nbd_export_aio_context(NBDExport *exp); 339 NBDExport *nbd_export_find(const char *name); 340 341 void nbd_client_new(QIOChannelSocket *sioc, 342 QCryptoTLSCreds *tlscreds, 343 const char *tlsauthz, 344 void (*close_fn)(NBDClient *, bool)); 345 void nbd_client_get(NBDClient *client); 346 void nbd_client_put(NBDClient *client); 347 348 void nbd_server_is_qemu_nbd(int max_connections); 349 bool nbd_server_is_running(void); 350 int nbd_server_max_connections(void); 351 void nbd_server_start(SocketAddress *addr, const char *tls_creds, 352 const char *tls_authz, uint32_t max_connections, 353 Error **errp); 354 void nbd_server_start_options(NbdServerOptions *arg, Error **errp); 355 356 /* nbd_read 357 * Reads @size bytes from @ioc. Returns 0 on success. 358 */ 359 static inline int nbd_read(QIOChannel *ioc, void *buffer, size_t size, 360 const char *desc, Error **errp) 361 { 362 ERRP_GUARD(); 363 int ret = qio_channel_read_all(ioc, buffer, size, errp) < 0 ? -EIO : 0; 364 365 if (ret < 0) { 366 if (desc) { 367 error_prepend(errp, "Failed to read %s: ", desc); 368 } 369 return ret; 370 } 371 372 return 0; 373 } 374 375 #define DEF_NBD_READ_N(bits) \ 376 static inline int nbd_read##bits(QIOChannel *ioc, \ 377 uint##bits##_t *val, \ 378 const char *desc, Error **errp) \ 379 { \ 380 int ret = nbd_read(ioc, val, sizeof(*val), desc, errp); \ 381 if (ret < 0) { \ 382 return ret; \ 383 } \ 384 *val = be##bits##_to_cpu(*val); \ 385 return 0; \ 386 } 387 388 DEF_NBD_READ_N(16) /* Defines nbd_read16(). */ 389 DEF_NBD_READ_N(32) /* Defines nbd_read32(). */ 390 DEF_NBD_READ_N(64) /* Defines nbd_read64(). */ 391 392 #undef DEF_NBD_READ_N 393 394 static inline bool nbd_reply_is_simple(NBDReply *reply) 395 { 396 return reply->magic == NBD_SIMPLE_REPLY_MAGIC; 397 } 398 399 static inline bool nbd_reply_is_structured(NBDReply *reply) 400 { 401 return reply->magic == NBD_STRUCTURED_REPLY_MAGIC; 402 } 403 404 const char *nbd_reply_type_lookup(uint16_t type); 405 const char *nbd_opt_lookup(uint32_t opt); 406 const char *nbd_rep_lookup(uint32_t rep); 407 const char *nbd_info_lookup(uint16_t info); 408 const char *nbd_cmd_lookup(uint16_t info); 409 const char *nbd_err_lookup(int err); 410 411 /* nbd/client-connection.c */ 412 typedef struct NBDClientConnection NBDClientConnection; 413 414 void nbd_client_connection_enable_retry(NBDClientConnection *conn); 415 416 NBDClientConnection *nbd_client_connection_new(const SocketAddress *saddr, 417 bool do_negotiation, 418 const char *export_name, 419 const char *x_dirty_bitmap, 420 QCryptoTLSCreds *tlscreds, 421 const char *tlshostname); 422 void nbd_client_connection_release(NBDClientConnection *conn); 423 424 QIOChannel *coroutine_fn 425 nbd_co_establish_connection(NBDClientConnection *conn, NBDExportInfo *info, 426 bool blocking, Error **errp); 427 428 void nbd_co_establish_connection_cancel(NBDClientConnection *conn); 429 430 #endif 431