]> git.proxmox.com Git - mirror_qemu.git/blob - docs/tools/qemu-nbd.rst
nbd/server: Allow MULTI_CONN for shared writable exports
[mirror_qemu.git] / docs / tools / qemu-nbd.rst
1 =====================================
2 QEMU Disk Network Block Device Server
3 =====================================
4
5 Synopsis
6 --------
7
8 **qemu-nbd** [*OPTION*]... *filename*
9
10 **qemu-nbd** -L [*OPTION*]...
11
12 **qemu-nbd** -d *dev*
13
14 Description
15 -----------
16
17 Export a QEMU disk image using the NBD protocol.
18
19 Other uses:
20
21 - Bind a /dev/nbdX block device to a QEMU server (on Linux).
22 - As a client to query exports of a remote NBD server.
23
24 Options
25 -------
26
27 .. program:: qemu-nbd
28
29 *filename* is a disk image filename, or a set of block
30 driver options if :option:`--image-opts` is specified.
31
32 *dev* is an NBD device.
33
34 .. option:: --object type,id=ID,...
35
36 Define a new instance of the *type* object class identified by *ID*.
37 See the :manpage:`qemu(1)` manual page for full details of the properties
38 supported. The common object types that it makes sense to define are the
39 ``secret`` object, which is used to supply passwords and/or encryption
40 keys, and the ``tls-creds`` object, which is used to supply TLS
41 credentials for the ``qemu-nbd`` server or client.
42
43 .. option:: -p, --port=PORT
44
45 TCP port to listen on as a server, or connect to as a client
46 (default ``10809``).
47
48 .. option:: -o, --offset=OFFSET
49
50 The offset into the image.
51
52 .. option:: -b, --bind=IFACE
53
54 The interface to bind to as a server, or connect to as a client
55 (default ``0.0.0.0``).
56
57 .. option:: -k, --socket=PATH
58
59 Use a unix socket with path *PATH*.
60
61 .. option:: --image-opts
62
63 Treat *filename* as a set of image options, instead of a plain
64 filename. If this flag is specified, the ``-f`` flag should
65 not be used, instead the :option:`format=` option should be set.
66
67 .. option:: -f, --format=FMT
68
69 Force the use of the block driver for format *FMT* instead of
70 auto-detecting.
71
72 .. option:: -r, --read-only
73
74 Export the disk as read-only.
75
76 .. option:: -A, --allocation-depth
77
78 Expose allocation depth information via the
79 ``qemu:allocation-depth`` metadata context accessible through
80 NBD_OPT_SET_META_CONTEXT.
81
82 .. option:: -B, --bitmap=NAME
83
84 If *filename* has a qcow2 persistent bitmap *NAME*, expose
85 that bitmap via the ``qemu:dirty-bitmap:NAME`` metadata context
86 accessible through NBD_OPT_SET_META_CONTEXT.
87
88 .. option:: -s, --snapshot
89
90 Use *filename* as an external snapshot, create a temporary
91 file with ``backing_file=``\ *filename*, redirect the write to
92 the temporary one.
93
94 .. option:: -l, --load-snapshot=SNAPSHOT_PARAM
95
96 Load an internal snapshot inside *filename* and export it
97 as an read-only device, SNAPSHOT_PARAM format is
98 ``snapshot.id=[ID],snapshot.name=[NAME]`` or ``[ID_OR_NAME]``
99
100 .. option:: --cache=CACHE
101
102 The cache mode to be used with the file. Valid values are:
103 ``none``, ``writeback`` (the default), ``writethrough``,
104 ``directsync`` and ``unsafe``. See the documentation of
105 the emulator's ``-drive cache=...`` option for more info.
106
107 .. option:: -n, --nocache
108
109 Equivalent to :option:`--cache=none`.
110
111 .. option:: --aio=AIO
112
113 Set the asynchronous I/O mode between ``threads`` (the default),
114 ``native`` (Linux only), and ``io_uring`` (Linux 5.1+).
115
116 .. option:: --discard=DISCARD
117
118 Control whether ``discard`` (also known as ``trim`` or ``unmap``)
119 requests are ignored or passed to the filesystem. *DISCARD* is one of
120 ``ignore`` (or ``off``), ``unmap`` (or ``on``). The default is
121 ``ignore``.
122
123 .. option:: --detect-zeroes=DETECT_ZEROES
124
125 Control the automatic conversion of plain zero writes by the OS to
126 driver-specific optimized zero write commands. *DETECT_ZEROES* is one of
127 ``off``, ``on``, or ``unmap``. ``unmap``
128 converts a zero write to an unmap operation and can only be used if
129 *DISCARD* is set to ``unmap``. The default is ``off``.
130
131 .. option:: -c, --connect=DEV
132
133 Connect *filename* to NBD device *DEV* (Linux only).
134
135 .. option:: -d, --disconnect
136
137 Disconnect the device *DEV* (Linux only).
138
139 .. option:: -e, --shared=NUM
140
141 Allow up to *NUM* clients to share the device (default
142 ``1``), 0 for unlimited.
143
144 .. option:: -t, --persistent
145
146 Don't exit on the last connection.
147
148 .. option:: -x, --export-name=NAME
149
150 Set the NBD volume export name (default of a zero-length string).
151
152 .. option:: -D, --description=DESCRIPTION
153
154 Set the NBD volume export description, as a human-readable
155 string.
156
157 .. option:: -L, --list
158
159 Connect as a client and list all details about the exports exposed by
160 a remote NBD server. This enables list mode, and is incompatible
161 with options that change behavior related to a specific export (such as
162 :option:`--export-name`, :option:`--offset`, ...).
163
164 .. option:: --tls-creds=ID
165
166 Enable mandatory TLS encryption for the server by setting the ID
167 of the TLS credentials object previously created with the
168 :option:`--object` option; or provide the credentials needed for
169 connecting as a client in list mode.
170
171 .. option:: --tls-hostname=hostname
172
173 When validating an x509 certificate received over a TLS connection,
174 the hostname that the NBD client used to connect will be checked
175 against information in the server provided certificate. Sometimes
176 it might be required to override the hostname used to perform this
177 check. For example, if the NBD client is using a tunnel from localhost
178 to connect to the remote server, the :option:`--tls-hostname` option should
179 be used to set the officially expected hostname of the remote NBD
180 server. This can also be used if accessing NBD over a UNIX socket
181 where there is no inherent hostname available. This is only permitted
182 when acting as a NBD client with the :option:`--list` option.
183
184 .. option:: --fork
185
186 Fork off the server process and exit the parent once the server is running.
187
188 .. option:: --pid-file=PATH
189
190 Store the server's process ID in the given file.
191
192 .. option:: --tls-authz=ID
193
194 Specify the ID of a qauthz object previously created with the
195 :option:`--object` option. This will be used to authorize connecting users
196 against their x509 distinguished name.
197
198 .. option:: -v, --verbose
199
200 Display extra debugging information.
201
202 .. option:: -h, --help
203
204 Display this help and exit.
205
206 .. option:: -V, --version
207
208 Display version information and exit.
209
210 .. option:: -T, --trace [[enable=]PATTERN][,events=FILE][,file=FILE]
211
212 .. include:: ../qemu-option-trace.rst.inc
213
214 Examples
215 --------
216
217 Start a server listening on port 10809 that exposes only the
218 guest-visible contents of a qcow2 file, with no TLS encryption, and
219 with the default export name (an empty string). The command is
220 one-shot, and will block until the first successful client
221 disconnects:
222
223 ::
224
225 qemu-nbd -f qcow2 file.qcow2
226
227 Start a long-running server listening with encryption on port 10810,
228 and whitelist clients with a specific X.509 certificate to connect to
229 a 1 megabyte subset of a raw file, using the export name 'subset':
230
231 ::
232
233 qemu-nbd \
234 --object tls-creds-x509,id=tls0,endpoint=server,dir=/path/to/qemutls \
235 --object 'authz-simple,id=auth0,identity=CN=laptop.example.com,,\
236 O=Example Org,,L=London,,ST=London,,C=GB' \
237 --tls-creds tls0 --tls-authz auth0 \
238 -t -x subset -p 10810 \
239 --image-opts driver=raw,offset=1M,size=1M,file.driver=file,file.filename=file.raw
240
241 Serve a read-only copy of a guest image over a Unix socket with as
242 many as 5 simultaneous readers, with a persistent process forked as a
243 daemon:
244
245 ::
246
247 qemu-nbd --fork --persistent --shared=5 --socket=/path/to/sock \
248 --read-only --format=qcow2 file.qcow2
249
250 Expose the guest-visible contents of a qcow2 file via a block device
251 /dev/nbd0 (and possibly creating /dev/nbd0p1 and friends for
252 partitions found within), then disconnect the device when done.
253 Access to bind ``qemu-nbd`` to a /dev/nbd device generally requires root
254 privileges, and may also require the execution of ``modprobe nbd``
255 to enable the kernel NBD client module. *CAUTION*: Do not use
256 this method to mount filesystems from an untrusted guest image - a
257 malicious guest may have prepared the image to attempt to trigger
258 kernel bugs in partition probing or file system mounting.
259
260 ::
261
262 qemu-nbd -c /dev/nbd0 -f qcow2 file.qcow2
263 qemu-nbd -d /dev/nbd0
264
265 Query a remote server to see details about what export(s) it is
266 serving on port 10809, and authenticating via PSK:
267
268 ::
269
270 qemu-nbd \
271 --object tls-creds-psk,id=tls0,dir=/tmp/keys,username=eblake,endpoint=client \
272 --tls-creds tls0 -L -b remote.example.com
273
274 See also
275 --------
276
277 :manpage:`qemu(1)`, :manpage:`qemu-img(1)`