3 * Copyright (C) 2005 Anthony Liguori <anthony@codemonkey.ws>
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.
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.
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/>.
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"
29 typedef struct NBDExport NBDExport
;
30 typedef struct NBDClient NBDClient
;
31 typedef struct NBDClientConnection NBDClientConnection
;
33 extern const BlockExportDriver blk_exp_nbd
;
35 /* Handshake phase structs - this struct is passed on the wire */
37 typedef struct NBDOption
{
38 uint64_t magic
; /* NBD_OPTS_MAGIC */
39 uint32_t option
; /* NBD_OPT_* */
41 } QEMU_PACKED NBDOption
;
43 typedef struct NBDOptionReply
{
44 uint64_t magic
; /* NBD_REP_MAGIC */
45 uint32_t option
; /* NBD_OPT_* */
46 uint32_t type
; /* NBD_REP_* */
48 } QEMU_PACKED NBDOptionReply
;
50 typedef struct NBDOptionReplyMetaContext
{
51 NBDOptionReply h
; /* h.type = NBD_REP_META_CONTEXT, h.length > 4 */
53 /* metadata context name follows */
54 } QEMU_PACKED NBDOptionReplyMetaContext
;
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 */
66 /* Transmission phase structs */
69 * Note: NBDRequest is _NOT_ the same as the network representation of an NBD
72 typedef struct NBDRequest
{
76 uint16_t flags
; /* NBD_CMD_FLAG_* */
77 uint16_t type
; /* NBD_CMD_* */
78 NBDMode mode
; /* Determines which network representation to use */
81 typedef struct NBDSimpleReply
{
82 uint32_t magic
; /* NBD_SIMPLE_REPLY_MAGIC */
85 } QEMU_PACKED NBDSimpleReply
;
87 /* Header of all structured replies */
88 typedef struct NBDStructuredReplyChunk
{
89 uint32_t magic
; /* NBD_STRUCTURED_REPLY_MAGIC */
90 uint16_t flags
; /* combination of NBD_REPLY_FLAG_* */
91 uint16_t type
; /* NBD_REPLY_TYPE_* */
92 uint64_t cookie
; /* request handle */
93 uint32_t length
; /* length of payload */
94 } QEMU_PACKED NBDStructuredReplyChunk
;
96 typedef union NBDReply
{
97 NBDSimpleReply simple
;
98 NBDStructuredReplyChunk structured
;
101 * @magic and @cookie fields have the same offset and size both in
102 * simple reply and structured reply chunk, so let them be accessible
103 * without ".simple." or ".structured." specification
111 /* Header of chunk for NBD_REPLY_TYPE_OFFSET_DATA */
112 typedef struct NBDStructuredReadData
{
113 /* header's .length >= 9 */
115 /* At least one byte of data payload follows, calculated from h.length */
116 } QEMU_PACKED NBDStructuredReadData
;
118 /* Complete chunk for NBD_REPLY_TYPE_OFFSET_HOLE */
119 typedef struct NBDStructuredReadHole
{
120 /* header's length == 12 */
123 } QEMU_PACKED NBDStructuredReadHole
;
125 /* Header of all NBD_REPLY_TYPE_ERROR* errors */
126 typedef struct NBDStructuredError
{
127 /* header's length >= 6 */
129 uint16_t message_length
;
130 } QEMU_PACKED NBDStructuredError
;
132 /* Header of NBD_REPLY_TYPE_BLOCK_STATUS */
133 typedef struct NBDStructuredMeta
{
134 /* header's length >= 12 (at least one extent) */
136 /* extents follows */
137 } QEMU_PACKED NBDStructuredMeta
;
139 /* Extent chunk for NBD_REPLY_TYPE_BLOCK_STATUS */
140 typedef struct NBDExtent
{
142 uint32_t flags
; /* NBD_STATE_* */
143 } QEMU_PACKED NBDExtent
;
145 /* Transmission (export) flags: sent from server to client during handshake,
146 but describe what will happen during transmission */
148 NBD_FLAG_HAS_FLAGS_BIT
= 0, /* Flags are there */
149 NBD_FLAG_READ_ONLY_BIT
= 1, /* Device is read-only */
150 NBD_FLAG_SEND_FLUSH_BIT
= 2, /* Send FLUSH */
151 NBD_FLAG_SEND_FUA_BIT
= 3, /* Send FUA (Force Unit Access) */
152 NBD_FLAG_ROTATIONAL_BIT
= 4, /* Use elevator algorithm -
154 NBD_FLAG_SEND_TRIM_BIT
= 5, /* Send TRIM (discard) */
155 NBD_FLAG_SEND_WRITE_ZEROES_BIT
= 6, /* Send WRITE_ZEROES */
156 NBD_FLAG_SEND_DF_BIT
= 7, /* Send DF (Do not Fragment) */
157 NBD_FLAG_CAN_MULTI_CONN_BIT
= 8, /* Multi-client cache consistent */
158 NBD_FLAG_SEND_RESIZE_BIT
= 9, /* Send resize */
159 NBD_FLAG_SEND_CACHE_BIT
= 10, /* Send CACHE (prefetch) */
160 NBD_FLAG_SEND_FAST_ZERO_BIT
= 11, /* FAST_ZERO flag for WRITE_ZEROES */
163 #define NBD_FLAG_HAS_FLAGS (1 << NBD_FLAG_HAS_FLAGS_BIT)
164 #define NBD_FLAG_READ_ONLY (1 << NBD_FLAG_READ_ONLY_BIT)
165 #define NBD_FLAG_SEND_FLUSH (1 << NBD_FLAG_SEND_FLUSH_BIT)
166 #define NBD_FLAG_SEND_FUA (1 << NBD_FLAG_SEND_FUA_BIT)
167 #define NBD_FLAG_ROTATIONAL (1 << NBD_FLAG_ROTATIONAL_BIT)
168 #define NBD_FLAG_SEND_TRIM (1 << NBD_FLAG_SEND_TRIM_BIT)
169 #define NBD_FLAG_SEND_WRITE_ZEROES (1 << NBD_FLAG_SEND_WRITE_ZEROES_BIT)
170 #define NBD_FLAG_SEND_DF (1 << NBD_FLAG_SEND_DF_BIT)
171 #define NBD_FLAG_CAN_MULTI_CONN (1 << NBD_FLAG_CAN_MULTI_CONN_BIT)
172 #define NBD_FLAG_SEND_RESIZE (1 << NBD_FLAG_SEND_RESIZE_BIT)
173 #define NBD_FLAG_SEND_CACHE (1 << NBD_FLAG_SEND_CACHE_BIT)
174 #define NBD_FLAG_SEND_FAST_ZERO (1 << NBD_FLAG_SEND_FAST_ZERO_BIT)
176 /* New-style handshake (global) flags, sent from server to client, and
177 control what will happen during handshake phase. */
178 #define NBD_FLAG_FIXED_NEWSTYLE (1 << 0) /* Fixed newstyle protocol. */
179 #define NBD_FLAG_NO_ZEROES (1 << 1) /* End handshake without zeroes. */
181 /* New-style client flags, sent from client to server to control what happens
182 during handshake phase. */
183 #define NBD_FLAG_C_FIXED_NEWSTYLE (1 << 0) /* Fixed newstyle protocol. */
184 #define NBD_FLAG_C_NO_ZEROES (1 << 1) /* End handshake without zeroes. */
186 /* Option requests. */
187 #define NBD_OPT_EXPORT_NAME (1)
188 #define NBD_OPT_ABORT (2)
189 #define NBD_OPT_LIST (3)
190 /* #define NBD_OPT_PEEK_EXPORT (4) not in use */
191 #define NBD_OPT_STARTTLS (5)
192 #define NBD_OPT_INFO (6)
193 #define NBD_OPT_GO (7)
194 #define NBD_OPT_STRUCTURED_REPLY (8)
195 #define NBD_OPT_LIST_META_CONTEXT (9)
196 #define NBD_OPT_SET_META_CONTEXT (10)
198 /* Option reply types. */
199 #define NBD_REP_ERR(value) ((UINT32_C(1) << 31) | (value))
201 #define NBD_REP_ACK (1) /* Data sending finished. */
202 #define NBD_REP_SERVER (2) /* Export description. */
203 #define NBD_REP_INFO (3) /* NBD_OPT_INFO/GO. */
204 #define NBD_REP_META_CONTEXT (4) /* NBD_OPT_{LIST,SET}_META_CONTEXT */
206 #define NBD_REP_ERR_UNSUP NBD_REP_ERR(1) /* Unknown option */
207 #define NBD_REP_ERR_POLICY NBD_REP_ERR(2) /* Server denied */
208 #define NBD_REP_ERR_INVALID NBD_REP_ERR(3) /* Invalid length */
209 #define NBD_REP_ERR_PLATFORM NBD_REP_ERR(4) /* Not compiled in */
210 #define NBD_REP_ERR_TLS_REQD NBD_REP_ERR(5) /* TLS required */
211 #define NBD_REP_ERR_UNKNOWN NBD_REP_ERR(6) /* Export unknown */
212 #define NBD_REP_ERR_SHUTDOWN NBD_REP_ERR(7) /* Server shutting down */
213 #define NBD_REP_ERR_BLOCK_SIZE_REQD NBD_REP_ERR(8) /* Need INFO_BLOCK_SIZE */
215 /* Info types, used during NBD_REP_INFO */
216 #define NBD_INFO_EXPORT 0
217 #define NBD_INFO_NAME 1
218 #define NBD_INFO_DESCRIPTION 2
219 #define NBD_INFO_BLOCK_SIZE 3
221 /* Request flags, sent from client to server during transmission phase */
222 #define NBD_CMD_FLAG_FUA (1 << 0) /* 'force unit access' during write */
223 #define NBD_CMD_FLAG_NO_HOLE (1 << 1) /* don't punch hole on zero run */
224 #define NBD_CMD_FLAG_DF (1 << 2) /* don't fragment structured read */
225 #define NBD_CMD_FLAG_REQ_ONE (1 << 3) /* only one extent in BLOCK_STATUS
227 #define NBD_CMD_FLAG_FAST_ZERO (1 << 4) /* fail if WRITE_ZEROES is not fast */
229 /* Supported request types */
237 NBD_CMD_WRITE_ZEROES
= 6,
238 NBD_CMD_BLOCK_STATUS
= 7,
241 #define NBD_DEFAULT_PORT 10809
243 /* Maximum size of a single READ/WRITE data buffer */
244 #define NBD_MAX_BUFFER_SIZE (32 * 1024 * 1024)
247 * Maximum size of a protocol string (export name, metadata context name,
248 * etc.). Use malloc rather than stack allocation for storage of a
251 #define NBD_MAX_STRING_SIZE 4096
253 /* Two types of reply structures */
254 #define NBD_SIMPLE_REPLY_MAGIC 0x67446698
255 #define NBD_STRUCTURED_REPLY_MAGIC 0x668e33ef
257 /* Structured reply flags */
258 #define NBD_REPLY_FLAG_DONE (1 << 0) /* This reply-chunk is last */
260 /* Structured reply types */
261 #define NBD_REPLY_ERR(value) ((1 << 15) | (value))
263 #define NBD_REPLY_TYPE_NONE 0
264 #define NBD_REPLY_TYPE_OFFSET_DATA 1
265 #define NBD_REPLY_TYPE_OFFSET_HOLE 2
266 #define NBD_REPLY_TYPE_BLOCK_STATUS 5
267 #define NBD_REPLY_TYPE_ERROR NBD_REPLY_ERR(1)
268 #define NBD_REPLY_TYPE_ERROR_OFFSET NBD_REPLY_ERR(2)
270 /* Extent flags for base:allocation in NBD_REPLY_TYPE_BLOCK_STATUS */
271 #define NBD_STATE_HOLE (1 << 0)
272 #define NBD_STATE_ZERO (1 << 1)
274 /* Extent flags for qemu:dirty-bitmap in NBD_REPLY_TYPE_BLOCK_STATUS */
275 #define NBD_STATE_DIRTY (1 << 0)
277 /* No flags needed for qemu:allocation-depth in NBD_REPLY_TYPE_BLOCK_STATUS */
279 static inline bool nbd_reply_type_is_error(int type
)
281 return type
& (1 << 15);
284 /* NBD errors are based on errno numbers, so there is a 1:1 mapping,
285 * but only a limited set of errno values is specified in the protocol.
286 * Everything else is squashed to EINVAL.
288 #define NBD_SUCCESS 0
291 #define NBD_ENOMEM 12
292 #define NBD_EINVAL 22
293 #define NBD_ENOSPC 28
294 #define NBD_EOVERFLOW 75
295 #define NBD_ENOTSUP 95
296 #define NBD_ESHUTDOWN 108
298 /* Details collected by NBD_OPT_EXPORT_NAME and NBD_OPT_GO */
299 typedef struct NBDExportInfo
{
300 /* Set by client before nbd_receive_negotiate() */
302 char *x_dirty_bitmap
;
304 /* Set by client before nbd_receive_negotiate(), or by server results
305 * during nbd_receive_export_list() */
306 char *name
; /* must be non-NULL */
308 /* In-out fields, set by client before nbd_receive_negotiate() and
309 * updated by server results during nbd_receive_negotiate() */
310 NBDMode mode
; /* input maximum mode tolerated; output actual mode chosen */
311 bool base_allocation
; /* base:allocation context for NBD_CMD_BLOCK_STATUS */
313 /* Set by server results during nbd_receive_negotiate() and
314 * nbd_receive_export_list() */
323 /* Set by server results during nbd_receive_export_list() */
329 int nbd_receive_negotiate(AioContext
*aio_context
, QIOChannel
*ioc
,
330 QCryptoTLSCreds
*tlscreds
,
331 const char *hostname
, QIOChannel
**outioc
,
332 NBDExportInfo
*info
, Error
**errp
);
333 void nbd_free_export_list(NBDExportInfo
*info
, int count
);
334 int nbd_receive_export_list(QIOChannel
*ioc
, QCryptoTLSCreds
*tlscreds
,
335 const char *hostname
, NBDExportInfo
**info
,
337 int nbd_init(int fd
, QIOChannelSocket
*sioc
, NBDExportInfo
*info
,
339 int nbd_send_request(QIOChannel
*ioc
, NBDRequest
*request
);
340 int coroutine_fn
nbd_receive_reply(BlockDriverState
*bs
, QIOChannel
*ioc
,
341 NBDReply
*reply
, Error
**errp
);
342 int nbd_client(int fd
);
343 int nbd_disconnect(int fd
);
344 int nbd_errno_to_system_errno(int err
);
346 void nbd_export_set_on_eject_blk(BlockExport
*exp
, BlockBackend
*blk
);
348 AioContext
*nbd_export_aio_context(NBDExport
*exp
);
349 NBDExport
*nbd_export_find(const char *name
);
351 void nbd_client_new(QIOChannelSocket
*sioc
,
352 QCryptoTLSCreds
*tlscreds
,
353 const char *tlsauthz
,
354 void (*close_fn
)(NBDClient
*, bool));
355 void nbd_client_get(NBDClient
*client
);
356 void nbd_client_put(NBDClient
*client
);
358 void nbd_server_is_qemu_nbd(int max_connections
);
359 bool nbd_server_is_running(void);
360 int nbd_server_max_connections(void);
361 void nbd_server_start(SocketAddress
*addr
, const char *tls_creds
,
362 const char *tls_authz
, uint32_t max_connections
,
364 void nbd_server_start_options(NbdServerOptions
*arg
, Error
**errp
);
367 * Reads @size bytes from @ioc. Returns 0 on success.
369 static inline int nbd_read(QIOChannel
*ioc
, void *buffer
, size_t size
,
370 const char *desc
, Error
**errp
)
373 int ret
= qio_channel_read_all(ioc
, buffer
, size
, errp
) < 0 ? -EIO
: 0;
377 error_prepend(errp
, "Failed to read %s: ", desc
);
385 #define DEF_NBD_READ_N(bits) \
386 static inline int nbd_read##bits(QIOChannel *ioc, \
387 uint##bits##_t *val, \
388 const char *desc, Error **errp) \
390 int ret = nbd_read(ioc, val, sizeof(*val), desc, errp); \
394 *val = be##bits##_to_cpu(*val); \
398 DEF_NBD_READ_N(16) /* Defines nbd_read16(). */
399 DEF_NBD_READ_N(32) /* Defines nbd_read32(). */
400 DEF_NBD_READ_N(64) /* Defines nbd_read64(). */
402 #undef DEF_NBD_READ_N
404 static inline bool nbd_reply_is_simple(NBDReply
*reply
)
406 return reply
->magic
== NBD_SIMPLE_REPLY_MAGIC
;
409 static inline bool nbd_reply_is_structured(NBDReply
*reply
)
411 return reply
->magic
== NBD_STRUCTURED_REPLY_MAGIC
;
414 const char *nbd_reply_type_lookup(uint16_t type
);
415 const char *nbd_opt_lookup(uint32_t opt
);
416 const char *nbd_rep_lookup(uint32_t rep
);
417 const char *nbd_info_lookup(uint16_t info
);
418 const char *nbd_cmd_lookup(uint16_t info
);
419 const char *nbd_err_lookup(int err
);
420 const char *nbd_mode_lookup(NBDMode mode
);
422 /* nbd/client-connection.c */
423 void nbd_client_connection_enable_retry(NBDClientConnection
*conn
);
425 NBDClientConnection
*nbd_client_connection_new(const SocketAddress
*saddr
,
427 const char *export_name
,
428 const char *x_dirty_bitmap
,
429 QCryptoTLSCreds
*tlscreds
,
430 const char *tlshostname
);
431 void nbd_client_connection_release(NBDClientConnection
*conn
);
433 QIOChannel
*coroutine_fn
434 nbd_co_establish_connection(NBDClientConnection
*conn
, NBDExportInfo
*info
,
435 bool blocking
, Error
**errp
);
437 void nbd_co_establish_connection_cancel(NBDClientConnection
*conn
);