1 /* 2 * Copyright (C) 2016-2017 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 24 #include "qemu-common.h" 25 #include "qemu/option.h" 26 #include "io/channel-socket.h" 27 #include "crypto/tlscreds.h" 28 29 /* Handshake phase structs - this struct is passed on the wire */ 30 31 struct nbd_option { 32 uint64_t magic; /* NBD_OPTS_MAGIC */ 33 uint32_t option; /* NBD_OPT_* */ 34 uint32_t length; 35 } QEMU_PACKED; 36 typedef struct nbd_option nbd_option; 37 38 struct nbd_opt_reply { 39 uint64_t magic; /* NBD_REP_MAGIC */ 40 uint32_t option; /* NBD_OPT_* */ 41 uint32_t type; /* NBD_REP_* */ 42 uint32_t length; 43 } QEMU_PACKED; 44 typedef struct nbd_opt_reply nbd_opt_reply; 45 46 /* Transmission phase structs 47 * 48 * Note: these are _NOT_ the same as the network representation of an NBD 49 * request and reply! 50 */ 51 struct NBDRequest { 52 uint64_t handle; 53 uint64_t from; 54 uint32_t len; 55 uint16_t flags; /* NBD_CMD_FLAG_* */ 56 uint16_t type; /* NBD_CMD_* */ 57 }; 58 typedef struct NBDRequest NBDRequest; 59 60 typedef struct NBDSimpleReply { 61 uint32_t magic; /* NBD_SIMPLE_REPLY_MAGIC */ 62 uint32_t error; 63 uint64_t handle; 64 } QEMU_PACKED NBDSimpleReply; 65 66 /* Header of all structured replies */ 67 typedef struct NBDStructuredReplyChunk { 68 uint32_t magic; /* NBD_STRUCTURED_REPLY_MAGIC */ 69 uint16_t flags; /* combination of NBD_REPLY_FLAG_* */ 70 uint16_t type; /* NBD_REPLY_TYPE_* */ 71 uint64_t handle; /* request handle */ 72 uint32_t length; /* length of payload */ 73 } QEMU_PACKED NBDStructuredReplyChunk; 74 75 typedef union NBDReply { 76 NBDSimpleReply simple; 77 NBDStructuredReplyChunk structured; 78 struct { 79 /* @magic and @handle fields have the same offset and size both in 80 * simple reply and structured reply chunk, so let them be accessible 81 * without ".simple." or ".structured." specification 82 */ 83 uint32_t magic; 84 uint32_t _skip; 85 uint64_t handle; 86 } QEMU_PACKED; 87 } NBDReply; 88 89 /* Header of NBD_REPLY_TYPE_OFFSET_DATA, complete NBD_REPLY_TYPE_OFFSET_HOLE */ 90 typedef struct NBDStructuredRead { 91 NBDStructuredReplyChunk h; 92 uint64_t offset; 93 } QEMU_PACKED NBDStructuredRead; 94 95 /* Header of all NBD_REPLY_TYPE_ERROR* errors */ 96 typedef struct NBDStructuredError { 97 NBDStructuredReplyChunk h; 98 uint32_t error; 99 uint16_t message_length; 100 } QEMU_PACKED NBDStructuredError; 101 102 /* Transmission (export) flags: sent from server to client during handshake, 103 but describe what will happen during transmission */ 104 #define NBD_FLAG_HAS_FLAGS (1 << 0) /* Flags are there */ 105 #define NBD_FLAG_READ_ONLY (1 << 1) /* Device is read-only */ 106 #define NBD_FLAG_SEND_FLUSH (1 << 2) /* Send FLUSH */ 107 #define NBD_FLAG_SEND_FUA (1 << 3) /* Send FUA (Force Unit Access) */ 108 #define NBD_FLAG_ROTATIONAL (1 << 4) /* Use elevator algorithm - 109 rotational media */ 110 #define NBD_FLAG_SEND_TRIM (1 << 5) /* Send TRIM (discard) */ 111 #define NBD_FLAG_SEND_WRITE_ZEROES (1 << 6) /* Send WRITE_ZEROES */ 112 #define NBD_FLAG_SEND_DF (1 << 7) /* Send DF (Do not Fragment) */ 113 114 /* New-style handshake (global) flags, sent from server to client, and 115 control what will happen during handshake phase. */ 116 #define NBD_FLAG_FIXED_NEWSTYLE (1 << 0) /* Fixed newstyle protocol. */ 117 #define NBD_FLAG_NO_ZEROES (1 << 1) /* End handshake without zeroes. */ 118 119 /* New-style client flags, sent from client to server to control what happens 120 during handshake phase. */ 121 #define NBD_FLAG_C_FIXED_NEWSTYLE (1 << 0) /* Fixed newstyle protocol. */ 122 #define NBD_FLAG_C_NO_ZEROES (1 << 1) /* End handshake without zeroes. */ 123 124 /* Option requests. */ 125 #define NBD_OPT_EXPORT_NAME (1) 126 #define NBD_OPT_ABORT (2) 127 #define NBD_OPT_LIST (3) 128 /* #define NBD_OPT_PEEK_EXPORT (4) not in use */ 129 #define NBD_OPT_STARTTLS (5) 130 #define NBD_OPT_INFO (6) 131 #define NBD_OPT_GO (7) 132 #define NBD_OPT_STRUCTURED_REPLY (8) 133 134 /* Option reply types. */ 135 #define NBD_REP_ERR(value) ((UINT32_C(1) << 31) | (value)) 136 137 #define NBD_REP_ACK (1) /* Data sending finished. */ 138 #define NBD_REP_SERVER (2) /* Export description. */ 139 #define NBD_REP_INFO (3) /* NBD_OPT_INFO/GO. */ 140 141 #define NBD_REP_ERR_UNSUP NBD_REP_ERR(1) /* Unknown option */ 142 #define NBD_REP_ERR_POLICY NBD_REP_ERR(2) /* Server denied */ 143 #define NBD_REP_ERR_INVALID NBD_REP_ERR(3) /* Invalid length */ 144 #define NBD_REP_ERR_PLATFORM NBD_REP_ERR(4) /* Not compiled in */ 145 #define NBD_REP_ERR_TLS_REQD NBD_REP_ERR(5) /* TLS required */ 146 #define NBD_REP_ERR_UNKNOWN NBD_REP_ERR(6) /* Export unknown */ 147 #define NBD_REP_ERR_SHUTDOWN NBD_REP_ERR(7) /* Server shutting down */ 148 #define NBD_REP_ERR_BLOCK_SIZE_REQD NBD_REP_ERR(8) /* Need INFO_BLOCK_SIZE */ 149 150 /* Info types, used during NBD_REP_INFO */ 151 #define NBD_INFO_EXPORT 0 152 #define NBD_INFO_NAME 1 153 #define NBD_INFO_DESCRIPTION 2 154 #define NBD_INFO_BLOCK_SIZE 3 155 156 /* Request flags, sent from client to server during transmission phase */ 157 #define NBD_CMD_FLAG_FUA (1 << 0) /* 'force unit access' during write */ 158 #define NBD_CMD_FLAG_NO_HOLE (1 << 1) /* don't punch hole on zero run */ 159 #define NBD_CMD_FLAG_DF (1 << 2) /* don't fragment structured read */ 160 161 /* Supported request types */ 162 enum { 163 NBD_CMD_READ = 0, 164 NBD_CMD_WRITE = 1, 165 NBD_CMD_DISC = 2, 166 NBD_CMD_FLUSH = 3, 167 NBD_CMD_TRIM = 4, 168 /* 5 reserved for failed experiment NBD_CMD_CACHE */ 169 NBD_CMD_WRITE_ZEROES = 6, 170 }; 171 172 #define NBD_DEFAULT_PORT 10809 173 174 /* Maximum size of a single READ/WRITE data buffer */ 175 #define NBD_MAX_BUFFER_SIZE (32 * 1024 * 1024) 176 177 /* Maximum size of an export name. The NBD spec requires 256 and 178 * suggests that servers support up to 4096, but we stick to only the 179 * required size so that we can stack-allocate the names, and because 180 * going larger would require an audit of more code to make sure we 181 * aren't overflowing some other buffer. */ 182 #define NBD_MAX_NAME_SIZE 256 183 184 /* Two types of reply structures */ 185 #define NBD_SIMPLE_REPLY_MAGIC 0x67446698 186 #define NBD_STRUCTURED_REPLY_MAGIC 0x668e33ef 187 188 /* Structured reply flags */ 189 #define NBD_REPLY_FLAG_DONE (1 << 0) /* This reply-chunk is last */ 190 191 /* Structured reply types */ 192 #define NBD_REPLY_ERR(value) ((1 << 15) | (value)) 193 194 #define NBD_REPLY_TYPE_NONE 0 195 #define NBD_REPLY_TYPE_OFFSET_DATA 1 196 #define NBD_REPLY_TYPE_OFFSET_HOLE 2 197 #define NBD_REPLY_TYPE_ERROR NBD_REPLY_ERR(1) 198 #define NBD_REPLY_TYPE_ERROR_OFFSET NBD_REPLY_ERR(2) 199 200 static inline bool nbd_reply_type_is_error(int type) 201 { 202 return type & (1 << 15); 203 } 204 205 /* NBD errors are based on errno numbers, so there is a 1:1 mapping, 206 * but only a limited set of errno values is specified in the protocol. 207 * Everything else is squashed to EINVAL. 208 */ 209 #define NBD_SUCCESS 0 210 #define NBD_EPERM 1 211 #define NBD_EIO 5 212 #define NBD_ENOMEM 12 213 #define NBD_EINVAL 22 214 #define NBD_ENOSPC 28 215 #define NBD_EOVERFLOW 75 216 #define NBD_ESHUTDOWN 108 217 218 /* Details collected by NBD_OPT_EXPORT_NAME and NBD_OPT_GO */ 219 struct NBDExportInfo { 220 /* Set by client before nbd_receive_negotiate() */ 221 bool request_sizes; 222 223 /* In-out fields, set by client before nbd_receive_negotiate() and 224 * updated by server results during nbd_receive_negotiate() */ 225 bool structured_reply; 226 227 /* Set by server results during nbd_receive_negotiate() */ 228 uint64_t size; 229 uint16_t flags; 230 uint32_t min_block; 231 uint32_t opt_block; 232 uint32_t max_block; 233 }; 234 typedef struct NBDExportInfo NBDExportInfo; 235 236 int nbd_receive_negotiate(QIOChannel *ioc, const char *name, 237 QCryptoTLSCreds *tlscreds, const char *hostname, 238 QIOChannel **outioc, NBDExportInfo *info, 239 Error **errp); 240 int nbd_init(int fd, QIOChannelSocket *sioc, NBDExportInfo *info, 241 Error **errp); 242 int nbd_send_request(QIOChannel *ioc, NBDRequest *request); 243 int nbd_receive_reply(QIOChannel *ioc, NBDReply *reply, Error **errp); 244 int nbd_client(int fd); 245 int nbd_disconnect(int fd); 246 int nbd_errno_to_system_errno(int err); 247 248 typedef struct NBDExport NBDExport; 249 typedef struct NBDClient NBDClient; 250 251 NBDExport *nbd_export_new(BlockDriverState *bs, off_t dev_offset, off_t size, 252 uint16_t nbdflags, void (*close)(NBDExport *), 253 bool writethrough, BlockBackend *on_eject_blk, 254 Error **errp); 255 void nbd_export_close(NBDExport *exp); 256 void nbd_export_get(NBDExport *exp); 257 void nbd_export_put(NBDExport *exp); 258 259 BlockBackend *nbd_export_get_blockdev(NBDExport *exp); 260 261 NBDExport *nbd_export_find(const char *name); 262 void nbd_export_set_name(NBDExport *exp, const char *name); 263 void nbd_export_set_description(NBDExport *exp, const char *description); 264 void nbd_export_close_all(void); 265 266 void nbd_client_new(NBDExport *exp, 267 QIOChannelSocket *sioc, 268 QCryptoTLSCreds *tlscreds, 269 const char *tlsaclname, 270 void (*close_fn)(NBDClient *, bool)); 271 void nbd_client_get(NBDClient *client); 272 void nbd_client_put(NBDClient *client); 273 274 void nbd_server_start(SocketAddress *addr, const char *tls_creds, 275 Error **errp); 276 277 278 /* nbd_read 279 * Reads @size bytes from @ioc. Returns 0 on success. 280 */ 281 static inline int nbd_read(QIOChannel *ioc, void *buffer, size_t size, 282 Error **errp) 283 { 284 return qio_channel_read_all(ioc, buffer, size, errp) < 0 ? -EIO : 0; 285 } 286 287 static inline bool nbd_reply_is_simple(NBDReply *reply) 288 { 289 return reply->magic == NBD_SIMPLE_REPLY_MAGIC; 290 } 291 292 static inline bool nbd_reply_is_structured(NBDReply *reply) 293 { 294 return reply->magic == NBD_STRUCTURED_REPLY_MAGIC; 295 } 296 297 const char *nbd_reply_type_lookup(uint16_t type); 298 299 #endif 300