1 /* 2 * Copyright Red Hat 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 typedef struct NBDExport NBDExport; 30 typedef struct NBDClient NBDClient; 31 typedef struct NBDClientConnection NBDClientConnection; 32 33 extern const BlockExportDriver blk_exp_nbd; 34 35 /* Handshake phase structs - this struct is passed on the wire */ 36 37 typedef struct NBDOption { 38 uint64_t magic; /* NBD_OPTS_MAGIC */ 39 uint32_t option; /* NBD_OPT_* */ 40 uint32_t length; 41 } QEMU_PACKED NBDOption; 42 43 typedef struct NBDOptionReply { 44 uint64_t magic; /* NBD_REP_MAGIC */ 45 uint32_t option; /* NBD_OPT_* */ 46 uint32_t type; /* NBD_REP_* */ 47 uint32_t length; 48 } QEMU_PACKED NBDOptionReply; 49 50 typedef struct NBDOptionReplyMetaContext { 51 NBDOptionReply h; /* h.type = NBD_REP_META_CONTEXT, h.length > 4 */ 52 uint32_t context_id; 53 /* metadata context name follows */ 54 } QEMU_PACKED NBDOptionReplyMetaContext; 55 56 /* Track results of negotiation */ 57 typedef enum NBDMode { 58 /* Keep this list in a continuum of increasing features. */ 59 NBD_MODE_OLDSTYLE, /* server lacks newstyle negotiation */ 60 NBD_MODE_EXPORT_NAME, /* newstyle but only OPT_EXPORT_NAME safe */ 61 NBD_MODE_SIMPLE, /* newstyle but only simple replies */ 62 NBD_MODE_STRUCTURED, /* newstyle, structured replies enabled */ 63 /* TODO add NBD_MODE_EXTENDED */ 64 } NBDMode; 65 66 /* Transmission phase structs 67 * 68 * Note: these are _NOT_ the same as the network representation of an NBD 69 * request and reply! 70 */ 71 typedef struct NBDRequest { 72 uint64_t cookie; 73 uint64_t from; 74 uint32_t len; 75 uint16_t flags; /* NBD_CMD_FLAG_* */ 76 uint16_t type; /* NBD_CMD_* */ 77 } NBDRequest; 78 79 typedef struct NBDSimpleReply { 80 uint32_t magic; /* NBD_SIMPLE_REPLY_MAGIC */ 81 uint32_t error; 82 uint64_t cookie; 83 } QEMU_PACKED NBDSimpleReply; 84 85 /* Header of all structured replies */ 86 typedef struct NBDStructuredReplyChunk { 87 uint32_t magic; /* NBD_STRUCTURED_REPLY_MAGIC */ 88 uint16_t flags; /* combination of NBD_REPLY_FLAG_* */ 89 uint16_t type; /* NBD_REPLY_TYPE_* */ 90 uint64_t cookie; /* request handle */ 91 uint32_t length; /* length of payload */ 92 } QEMU_PACKED NBDStructuredReplyChunk; 93 94 typedef union NBDReply { 95 NBDSimpleReply simple; 96 NBDStructuredReplyChunk structured; 97 struct { 98 /* 99 * @magic and @cookie fields have the same offset and size both in 100 * simple reply and structured reply chunk, so let them be accessible 101 * without ".simple." or ".structured." specification 102 */ 103 uint32_t magic; 104 uint32_t _skip; 105 uint64_t cookie; 106 } QEMU_PACKED; 107 } NBDReply; 108 109 /* Header of chunk for NBD_REPLY_TYPE_OFFSET_DATA */ 110 typedef struct NBDStructuredReadData { 111 /* header's .length >= 9 */ 112 uint64_t offset; 113 /* At least one byte of data payload follows, calculated from h.length */ 114 } QEMU_PACKED NBDStructuredReadData; 115 116 /* Complete chunk for NBD_REPLY_TYPE_OFFSET_HOLE */ 117 typedef struct NBDStructuredReadHole { 118 /* header's length == 12 */ 119 uint64_t offset; 120 uint32_t length; 121 } QEMU_PACKED NBDStructuredReadHole; 122 123 /* Header of all NBD_REPLY_TYPE_ERROR* errors */ 124 typedef struct NBDStructuredError { 125 /* header's length >= 6 */ 126 uint32_t error; 127 uint16_t message_length; 128 } QEMU_PACKED NBDStructuredError; 129 130 /* Header of NBD_REPLY_TYPE_BLOCK_STATUS */ 131 typedef struct NBDStructuredMeta { 132 /* header's length >= 12 (at least one extent) */ 133 uint32_t context_id; 134 /* extents follows */ 135 } QEMU_PACKED NBDStructuredMeta; 136 137 /* Extent chunk for NBD_REPLY_TYPE_BLOCK_STATUS */ 138 typedef struct NBDExtent { 139 uint32_t length; 140 uint32_t flags; /* NBD_STATE_* */ 141 } QEMU_PACKED NBDExtent; 142 143 /* Transmission (export) flags: sent from server to client during handshake, 144 but describe what will happen during transmission */ 145 enum { 146 NBD_FLAG_HAS_FLAGS_BIT = 0, /* Flags are there */ 147 NBD_FLAG_READ_ONLY_BIT = 1, /* Device is read-only */ 148 NBD_FLAG_SEND_FLUSH_BIT = 2, /* Send FLUSH */ 149 NBD_FLAG_SEND_FUA_BIT = 3, /* Send FUA (Force Unit Access) */ 150 NBD_FLAG_ROTATIONAL_BIT = 4, /* Use elevator algorithm - 151 rotational media */ 152 NBD_FLAG_SEND_TRIM_BIT = 5, /* Send TRIM (discard) */ 153 NBD_FLAG_SEND_WRITE_ZEROES_BIT = 6, /* Send WRITE_ZEROES */ 154 NBD_FLAG_SEND_DF_BIT = 7, /* Send DF (Do not Fragment) */ 155 NBD_FLAG_CAN_MULTI_CONN_BIT = 8, /* Multi-client cache consistent */ 156 NBD_FLAG_SEND_RESIZE_BIT = 9, /* Send resize */ 157 NBD_FLAG_SEND_CACHE_BIT = 10, /* Send CACHE (prefetch) */ 158 NBD_FLAG_SEND_FAST_ZERO_BIT = 11, /* FAST_ZERO flag for WRITE_ZEROES */ 159 }; 160 161 #define NBD_FLAG_HAS_FLAGS (1 << NBD_FLAG_HAS_FLAGS_BIT) 162 #define NBD_FLAG_READ_ONLY (1 << NBD_FLAG_READ_ONLY_BIT) 163 #define NBD_FLAG_SEND_FLUSH (1 << NBD_FLAG_SEND_FLUSH_BIT) 164 #define NBD_FLAG_SEND_FUA (1 << NBD_FLAG_SEND_FUA_BIT) 165 #define NBD_FLAG_ROTATIONAL (1 << NBD_FLAG_ROTATIONAL_BIT) 166 #define NBD_FLAG_SEND_TRIM (1 << NBD_FLAG_SEND_TRIM_BIT) 167 #define NBD_FLAG_SEND_WRITE_ZEROES (1 << NBD_FLAG_SEND_WRITE_ZEROES_BIT) 168 #define NBD_FLAG_SEND_DF (1 << NBD_FLAG_SEND_DF_BIT) 169 #define NBD_FLAG_CAN_MULTI_CONN (1 << NBD_FLAG_CAN_MULTI_CONN_BIT) 170 #define NBD_FLAG_SEND_RESIZE (1 << NBD_FLAG_SEND_RESIZE_BIT) 171 #define NBD_FLAG_SEND_CACHE (1 << NBD_FLAG_SEND_CACHE_BIT) 172 #define NBD_FLAG_SEND_FAST_ZERO (1 << NBD_FLAG_SEND_FAST_ZERO_BIT) 173 174 /* New-style handshake (global) flags, sent from server to client, and 175 control what will happen during handshake phase. */ 176 #define NBD_FLAG_FIXED_NEWSTYLE (1 << 0) /* Fixed newstyle protocol. */ 177 #define NBD_FLAG_NO_ZEROES (1 << 1) /* End handshake without zeroes. */ 178 179 /* New-style client flags, sent from client to server to control what happens 180 during handshake phase. */ 181 #define NBD_FLAG_C_FIXED_NEWSTYLE (1 << 0) /* Fixed newstyle protocol. */ 182 #define NBD_FLAG_C_NO_ZEROES (1 << 1) /* End handshake without zeroes. */ 183 184 /* Option requests. */ 185 #define NBD_OPT_EXPORT_NAME (1) 186 #define NBD_OPT_ABORT (2) 187 #define NBD_OPT_LIST (3) 188 /* #define NBD_OPT_PEEK_EXPORT (4) not in use */ 189 #define NBD_OPT_STARTTLS (5) 190 #define NBD_OPT_INFO (6) 191 #define NBD_OPT_GO (7) 192 #define NBD_OPT_STRUCTURED_REPLY (8) 193 #define NBD_OPT_LIST_META_CONTEXT (9) 194 #define NBD_OPT_SET_META_CONTEXT (10) 195 196 /* Option reply types. */ 197 #define NBD_REP_ERR(value) ((UINT32_C(1) << 31) | (value)) 198 199 #define NBD_REP_ACK (1) /* Data sending finished. */ 200 #define NBD_REP_SERVER (2) /* Export description. */ 201 #define NBD_REP_INFO (3) /* NBD_OPT_INFO/GO. */ 202 #define NBD_REP_META_CONTEXT (4) /* NBD_OPT_{LIST,SET}_META_CONTEXT */ 203 204 #define NBD_REP_ERR_UNSUP NBD_REP_ERR(1) /* Unknown option */ 205 #define NBD_REP_ERR_POLICY NBD_REP_ERR(2) /* Server denied */ 206 #define NBD_REP_ERR_INVALID NBD_REP_ERR(3) /* Invalid length */ 207 #define NBD_REP_ERR_PLATFORM NBD_REP_ERR(4) /* Not compiled in */ 208 #define NBD_REP_ERR_TLS_REQD NBD_REP_ERR(5) /* TLS required */ 209 #define NBD_REP_ERR_UNKNOWN NBD_REP_ERR(6) /* Export unknown */ 210 #define NBD_REP_ERR_SHUTDOWN NBD_REP_ERR(7) /* Server shutting down */ 211 #define NBD_REP_ERR_BLOCK_SIZE_REQD NBD_REP_ERR(8) /* Need INFO_BLOCK_SIZE */ 212 213 /* Info types, used during NBD_REP_INFO */ 214 #define NBD_INFO_EXPORT 0 215 #define NBD_INFO_NAME 1 216 #define NBD_INFO_DESCRIPTION 2 217 #define NBD_INFO_BLOCK_SIZE 3 218 219 /* Request flags, sent from client to server during transmission phase */ 220 #define NBD_CMD_FLAG_FUA (1 << 0) /* 'force unit access' during write */ 221 #define NBD_CMD_FLAG_NO_HOLE (1 << 1) /* don't punch hole on zero run */ 222 #define NBD_CMD_FLAG_DF (1 << 2) /* don't fragment structured read */ 223 #define NBD_CMD_FLAG_REQ_ONE (1 << 3) /* only one extent in BLOCK_STATUS 224 * reply chunk */ 225 #define NBD_CMD_FLAG_FAST_ZERO (1 << 4) /* fail if WRITE_ZEROES is not fast */ 226 227 /* Supported request types */ 228 enum { 229 NBD_CMD_READ = 0, 230 NBD_CMD_WRITE = 1, 231 NBD_CMD_DISC = 2, 232 NBD_CMD_FLUSH = 3, 233 NBD_CMD_TRIM = 4, 234 NBD_CMD_CACHE = 5, 235 NBD_CMD_WRITE_ZEROES = 6, 236 NBD_CMD_BLOCK_STATUS = 7, 237 }; 238 239 #define NBD_DEFAULT_PORT 10809 240 241 /* Maximum size of a single READ/WRITE data buffer */ 242 #define NBD_MAX_BUFFER_SIZE (32 * 1024 * 1024) 243 244 /* 245 * Maximum size of a protocol string (export name, metadata context name, 246 * etc.). Use malloc rather than stack allocation for storage of a 247 * string. 248 */ 249 #define NBD_MAX_STRING_SIZE 4096 250 251 /* Two types of reply structures */ 252 #define NBD_SIMPLE_REPLY_MAGIC 0x67446698 253 #define NBD_STRUCTURED_REPLY_MAGIC 0x668e33ef 254 255 /* Structured reply flags */ 256 #define NBD_REPLY_FLAG_DONE (1 << 0) /* This reply-chunk is last */ 257 258 /* Structured reply types */ 259 #define NBD_REPLY_ERR(value) ((1 << 15) | (value)) 260 261 #define NBD_REPLY_TYPE_NONE 0 262 #define NBD_REPLY_TYPE_OFFSET_DATA 1 263 #define NBD_REPLY_TYPE_OFFSET_HOLE 2 264 #define NBD_REPLY_TYPE_BLOCK_STATUS 5 265 #define NBD_REPLY_TYPE_ERROR NBD_REPLY_ERR(1) 266 #define NBD_REPLY_TYPE_ERROR_OFFSET NBD_REPLY_ERR(2) 267 268 /* Extent flags for base:allocation in NBD_REPLY_TYPE_BLOCK_STATUS */ 269 #define NBD_STATE_HOLE (1 << 0) 270 #define NBD_STATE_ZERO (1 << 1) 271 272 /* Extent flags for qemu:dirty-bitmap in NBD_REPLY_TYPE_BLOCK_STATUS */ 273 #define NBD_STATE_DIRTY (1 << 0) 274 275 /* No flags needed for qemu:allocation-depth in NBD_REPLY_TYPE_BLOCK_STATUS */ 276 277 static inline bool nbd_reply_type_is_error(int type) 278 { 279 return type & (1 << 15); 280 } 281 282 /* NBD errors are based on errno numbers, so there is a 1:1 mapping, 283 * but only a limited set of errno values is specified in the protocol. 284 * Everything else is squashed to EINVAL. 285 */ 286 #define NBD_SUCCESS 0 287 #define NBD_EPERM 1 288 #define NBD_EIO 5 289 #define NBD_ENOMEM 12 290 #define NBD_EINVAL 22 291 #define NBD_ENOSPC 28 292 #define NBD_EOVERFLOW 75 293 #define NBD_ENOTSUP 95 294 #define NBD_ESHUTDOWN 108 295 296 /* Details collected by NBD_OPT_EXPORT_NAME and NBD_OPT_GO */ 297 typedef struct NBDExportInfo { 298 /* Set by client before nbd_receive_negotiate() */ 299 bool request_sizes; 300 char *x_dirty_bitmap; 301 302 /* Set by client before nbd_receive_negotiate(), or by server results 303 * during nbd_receive_export_list() */ 304 char *name; /* must be non-NULL */ 305 306 /* In-out fields, set by client before nbd_receive_negotiate() and 307 * updated by server results during nbd_receive_negotiate() */ 308 bool structured_reply; 309 bool base_allocation; /* base:allocation context for NBD_CMD_BLOCK_STATUS */ 310 311 /* Set by server results during nbd_receive_negotiate() and 312 * nbd_receive_export_list() */ 313 uint64_t size; 314 uint16_t flags; 315 uint32_t min_block; 316 uint32_t opt_block; 317 uint32_t max_block; 318 319 uint32_t context_id; 320 321 /* Set by server results during nbd_receive_export_list() */ 322 char *description; 323 int n_contexts; 324 char **contexts; 325 } NBDExportInfo; 326 327 int nbd_receive_negotiate(AioContext *aio_context, QIOChannel *ioc, 328 QCryptoTLSCreds *tlscreds, 329 const char *hostname, QIOChannel **outioc, 330 NBDExportInfo *info, Error **errp); 331 void nbd_free_export_list(NBDExportInfo *info, int count); 332 int nbd_receive_export_list(QIOChannel *ioc, QCryptoTLSCreds *tlscreds, 333 const char *hostname, NBDExportInfo **info, 334 Error **errp); 335 int nbd_init(int fd, QIOChannelSocket *sioc, NBDExportInfo *info, 336 Error **errp); 337 int nbd_send_request(QIOChannel *ioc, NBDRequest *request); 338 int coroutine_fn nbd_receive_reply(BlockDriverState *bs, QIOChannel *ioc, 339 NBDReply *reply, Error **errp); 340 int nbd_client(int fd); 341 int nbd_disconnect(int fd); 342 int nbd_errno_to_system_errno(int err); 343 344 void nbd_export_set_on_eject_blk(BlockExport *exp, BlockBackend *blk); 345 346 AioContext *nbd_export_aio_context(NBDExport *exp); 347 NBDExport *nbd_export_find(const char *name); 348 349 void nbd_client_new(QIOChannelSocket *sioc, 350 QCryptoTLSCreds *tlscreds, 351 const char *tlsauthz, 352 void (*close_fn)(NBDClient *, bool)); 353 void nbd_client_get(NBDClient *client); 354 void nbd_client_put(NBDClient *client); 355 356 void nbd_server_is_qemu_nbd(int max_connections); 357 bool nbd_server_is_running(void); 358 int nbd_server_max_connections(void); 359 void nbd_server_start(SocketAddress *addr, const char *tls_creds, 360 const char *tls_authz, uint32_t max_connections, 361 Error **errp); 362 void nbd_server_start_options(NbdServerOptions *arg, Error **errp); 363 364 /* nbd_read 365 * Reads @size bytes from @ioc. Returns 0 on success. 366 */ 367 static inline int nbd_read(QIOChannel *ioc, void *buffer, size_t size, 368 const char *desc, Error **errp) 369 { 370 ERRP_GUARD(); 371 int ret = qio_channel_read_all(ioc, buffer, size, errp) < 0 ? -EIO : 0; 372 373 if (ret < 0) { 374 if (desc) { 375 error_prepend(errp, "Failed to read %s: ", desc); 376 } 377 return ret; 378 } 379 380 return 0; 381 } 382 383 #define DEF_NBD_READ_N(bits) \ 384 static inline int nbd_read##bits(QIOChannel *ioc, \ 385 uint##bits##_t *val, \ 386 const char *desc, Error **errp) \ 387 { \ 388 int ret = nbd_read(ioc, val, sizeof(*val), desc, errp); \ 389 if (ret < 0) { \ 390 return ret; \ 391 } \ 392 *val = be##bits##_to_cpu(*val); \ 393 return 0; \ 394 } 395 396 DEF_NBD_READ_N(16) /* Defines nbd_read16(). */ 397 DEF_NBD_READ_N(32) /* Defines nbd_read32(). */ 398 DEF_NBD_READ_N(64) /* Defines nbd_read64(). */ 399 400 #undef DEF_NBD_READ_N 401 402 static inline bool nbd_reply_is_simple(NBDReply *reply) 403 { 404 return reply->magic == NBD_SIMPLE_REPLY_MAGIC; 405 } 406 407 static inline bool nbd_reply_is_structured(NBDReply *reply) 408 { 409 return reply->magic == NBD_STRUCTURED_REPLY_MAGIC; 410 } 411 412 const char *nbd_reply_type_lookup(uint16_t type); 413 const char *nbd_opt_lookup(uint32_t opt); 414 const char *nbd_rep_lookup(uint32_t rep); 415 const char *nbd_info_lookup(uint16_t info); 416 const char *nbd_cmd_lookup(uint16_t info); 417 const char *nbd_err_lookup(int err); 418 const char *nbd_mode_lookup(NBDMode mode); 419 420 /* nbd/client-connection.c */ 421 void nbd_client_connection_enable_retry(NBDClientConnection *conn); 422 423 NBDClientConnection *nbd_client_connection_new(const SocketAddress *saddr, 424 bool do_negotiation, 425 const char *export_name, 426 const char *x_dirty_bitmap, 427 QCryptoTLSCreds *tlscreds, 428 const char *tlshostname); 429 void nbd_client_connection_release(NBDClientConnection *conn); 430 431 QIOChannel *coroutine_fn 432 nbd_co_establish_connection(NBDClientConnection *conn, NBDExportInfo *info, 433 bool blocking, Error **errp); 434 435 void nbd_co_establish_connection_cancel(NBDClientConnection *conn); 436 437 #endif 438