3 nbdkit - which parts of the NBD protocol nbdkit supports
7 nbdkit [-n|--newstyle] [--mask-handshake MASK] [--no-sr] [-o|--oldstyle]
12 This page documents the level of support in nbdkit for various parts
15 =head1 NEW STYLE VS OLD STYLE PROTOCOL
17 The NBD protocol comes in two incompatible forms that we call
18 "oldstyle" and "newstyle". Unfortunately which protocol you should
19 use depends on the client and cannot be known in advance, nor can it
20 be negotiated from the server side.
22 nbdkit defaults to the newstyle protocol since nbdkit E<ge> 1.3. The
23 newstyle protocol is better in every respect than the oldstyle
24 protocol and you should prefer it if possible. The newstyle protocol
25 also includes an extension where a client may request structured
26 replies for even more capabilities, such as sparse reads or obtaining
27 block status. By default, nbdkit advertises as many features as it
28 can support (in some cases, this can be limited by what callbacks the
29 plugin handles), even if the client does not negotiate to use all
32 Nbdkit also includes some options that are useful mainly when
33 performing integration tests, for proving whether clients have sane
34 fallback behavior when dealing various older servers permitted by the
35 NBD protocol. Use the I<--no-sr> flag to force the newstyle protocol
36 to decline any client request for structured replies. Use the
37 I<--mask-handshake> parameter to mask off particular global features
38 which are advertised during new-style handshake (defaulting to all
39 supported bits set). Clearing bit 0 (the low order bit) limits a
40 client to using just C<NBD_OPT_EXPORT_NAME> (and is incompatible with
41 TLS or structured replies); clearing bit 1 causes the handshake to
42 send more padding bytes in response to C<NBD_OPT_EXPORT_NAME>. Other
43 bits in the mask will only have an effect if the NBD protocol is
44 extended in the future to define other global bits.
46 Use the I<-o> or I<--oldstyle> flag to force the oldstyle protocol.
47 In this mode, I<--no-sr> and I<--mask-handshake> have no effect.
49 =head2 Common clients and the protocol they require
52 ------------------------------------------------------------
53 qemu <= 2.5 without exportname oldstyle
54 qemu <= 2.5 with exportname newstyle
55 qemu >= 2.6 client can talk either protocol
56 qemu >= 2.11 client tries structured replies
57 nbd-client < 3.10 client can talk either protocol
58 nbd-client >= 3.10 newstyle, no structured replies
59 any TLS (encrypted) client newstyle
60 nbdkit nbd plugin client can talk either protocol
61 nbdkit >= 1.13.3 nbd plugin tries structured replies
62 libnbd either protocol, tries structured replies
64 =head2 Errors seen if using the wrong protocol
66 If you use qemu E<le> 2.5 without the exportname field against a
67 newstyle server, it will give the error:
69 Server requires an export name
71 If you use qemu E<le> 2.5 with the exportname field against an
72 oldstyle server, it will give the error:
74 Server does not support export names
76 If you use the oldstyle protocol with nbd-client E<ge> 3.10, it will
79 Error: It looks like you're trying to connect to an oldstyle server.
81 =head2 NBD protocol and port number
83 Port 10809/tcp is reserved by IANA for the NBD protocol, but you can
84 use nbdkit on any port or on Unix domain sockets.
86 The NBD protocol specification claims that you should always use
87 newstyle when using port 10809, and use oldstyle on all other ports,
88 but this claim is not based on the reality of what NBD servers do, and
89 nbdkit does not require or encourage this.
91 =head1 NBD PROTOCOL FEATURES SUPPORTED BY NBDKIT
95 =item newstyle protocol
97 Supported in nbdkit E<ge> 1.1.12, and the default in nbdkit E<ge> 1.3.
101 Partially supported in nbdkit E<ge> 1.1.12. Support for plugins to
102 read the client export name added in nbdkit E<ge> 1.15.2.
104 Versions of nbdkit before 1.16 could advertise a single export name to
105 clients, via a now deprecated side effect of the I<-e> option. In nbdkit
106 1.15.2, plugins could read the client requested export name using
107 C<nbdkit_export_name()> and serve different content. In nbdkit
108 1.21.22, plugins could implement C<.list_exports> to answer
109 C<NBD_OPT_LIST> queries.
111 =item C<NBD_FLAG_NO_ZEROES>
113 Supported in nbdkit E<ge> 1.1.13.
115 This protocol optimization avoids sending a useless block of zero
116 bytes during protocol negotiation.
118 =item C<NBD_CMD_WRITE_ZEROES>
120 Supported in nbdkit E<ge> 1.1.13.
122 =item TLS with X.509 certificates
124 Supported in nbdkit E<ge> 1.1.15.
126 =item TLS with Pre-Shared Keys (PSK)
128 Supported in nbdkit E<ge> 1.4.0.
132 Supported in nbdkit E<ge> 1.5.3.
134 This protocol enhancement allows the server to return errors when
135 negotiating the export name.
137 =item C<NBD_OPT_INFO>
139 Supported in nbdkit E<ge> 1.9.3.
141 This protocol enhancement allows a client to inspect details about
142 the export without actually connecting.
144 =item C<NBD_FLAG_CAN_MULTI_CONN>
146 Supported in nbdkit E<ge> 1.9.9.
148 =item Structured Replies
150 Supported in nbdkit E<ge> 1.11.8.
152 However we don’t expose the capability to send structured replies to
153 plugins yet, nor do we send human-readable error messages using this
156 In nbdkit E<ge> 1.13.9>, the command-line option I<--no-sr> can be
157 used to disable server support for structured replies, for testing
160 =item Metadata Querying
162 Supported in nbdkit E<ge> 1.11.8.
166 Supported in nbdkit E<ge> 1.11.10.
168 Only C<base:allocation> (ie. querying which parts of an image are
169 sparse) is supported.
171 Sparse reads (using C<NBD_REPLY_TYPE_OFFSET_HOLE> are not directly
172 supported, but a client can use block status to infer which portions
173 of the export do not need to be read.
177 Supported in nbdkit E<ge> 1.11.11.
179 This protocol extension allows a client to force an all-or-none read
180 when structured replies are in effect. However, the flag is a no-op
181 until we extend the plugin API to allow a fragmented read in the first
184 =item C<NBD_CMD_CACHE>
186 Supported in nbdkit E<ge> 1.13.4.
188 This protocol extension allows a client to inform the server about
189 intent to access a portion of the export, to allow the server an
190 opportunity to cache things appropriately.
192 =item C<NBD_CMD_FLAG_FAST_ZERO>
194 Supported in nbdkit E<ge> 1.15.0.
196 This protocol extension allows a server to advertise that it can rank
197 all zero requests as fast or slow, at which point the client can make
198 fast zero requests which fail immediately with C<ENOTSUP> if the
199 request is no faster than a counterpart write would be, while normal
200 zero requests still benefit from compressed network traffic regardless
203 =item C<NBD_INFO_NAME>, C<NBD_INFO_DESCRIPTION>
205 Supported in nbdkit E<ge> 1.23.6.
207 These protocol extensions allow a client to learn more information
208 about an export during C<NBD_OPT_GO>. The C<.default_export> callback
209 can inform a client of a canonical non-empty name in place of the
210 default export C<"">, and the C<.export_description> callback can give
211 a client more details about the export.
213 =item C<NBD_INFO_BLOCK_SIZE>
215 Supported in nbdkit E<ge> 1.30.
217 =item Resize Extension
226 L<https://github.com/NetworkBlockDevice/nbd/blob/master/doc/proto.md>,
227 L<https://nbd.sourceforge.io/>.
239 Copyright (C) 2013-2020 Red Hat Inc.