NAME

nbdkit - which parts of the NBD protocol nbdkit supports

SYNOPSIS

 nbdkit [-n|--newstyle] [--mask-handshake MASK] [--no-sr] [-o|--oldstyle]
        [-e|--exportname EXPORTNAME] [...]

DESCRIPTION

This page documents the level of support in nbdkit for various parts of the NBD protocol.

NEW STYLE VS OLD STYLE PROTOCOL

The NBD protocol comes in two incompatible forms that we call "oldstyle" and "newstyle". Unfortunately which protocol you should use depends on the client and cannot be known in advance, nor can it be negotiated from the server side.

nbdkit defaults to the newstyle protocol since nbdkit ≥ 1.3. The newstyle protocol is better in every respect than the oldstyle protocol and you should prefer it if possible. The newstyle protocol also includes an extension where a client may request structured replies for even more capabilities, such as sparse reads or obtaining block status. By default, nbdkit advertises as many features as it can support (in some cases, this can be limited by what callbacks the plugin handles), even if the client does not negotiate to use all advertised features.

Use the -e or --exportname flag to set the optional exportname for the newstyle protocol.

Nbdkit also includes some options that are useful mainly when performing integration tests, for proving whether clients have sane fallback behavior when dealing various older servers permitted by the NBD protocol. Use the --no-sr flag to force the newstyle protocol to decline any client request for structured replies. Use the --mask-handshake parameter to mask off particular global features which are advertised during new-style handshake (defaulting to all supported bits set). Clearing bit 0 (the low order bit) limits a client to using just "NBD_OPT_EXPORT_NAME" (and is incompatible with TLS or structured replies); clearing bit 1 causes the handshake to send more padding bytes in response to "NBD_OPT_EXPORT_NAME". Other bits in the mask will only have an effect if the NBD protocol is extended in the future to define other global bits.

Use the -o or --oldstyle flag to force the oldstyle protocol. In this mode, --no-sr and --mask-handshake have no effect.

Common clients and the protocol they require

 Client                          Protocol
 ------------------------------------------------------------
 qemu <= 2.5 without exportname  oldstyle
 qemu <= 2.5 with exportname     newstyle
 qemu >= 2.6                     client can talk either protocol
 qemu >= 2.11                    client tries structured replies
 nbd-client < 3.10               client can talk either protocol
 nbd-client >= 3.10              newstyle, no structured replies
 any TLS (encrypted) client      newstyle
 nbdkit nbd plugin               client can talk either protocol
 nbdkit >= 1.13.3                nbd plugin tries structured replies
 libnbd                          either protocol, tries structured replies

Errors seen if using the wrong protocol

If you use qemu ≤ 2.5 without the exportname field against a newstyle server, it will give the error:

 Server requires an export name

If you use qemu ≤ 2.5 with the exportname field against an oldstyle server, it will give the error:

 Server does not support export names

If you use the oldstyle protocol with nbd-client ≥ 3.10, it will give the error:

 Error: It looks like you're trying to connect to an oldstyle server.

NBD protocol and port number

Port 10809/tcp is reserved by IANA for the NBD protocol, but you can use nbdkit on any port or on Unix domain sockets.

The NBD protocol specification claims that you should always use newstyle when using port 10809, and use oldstyle on all other ports, but this claim is not based on the reality of what NBD servers do, and nbdkit does not require or encourage this.

NBD PROTOCOL FEATURES SUPPORTED BY NBDKIT

newstyle protocol

Supported in nbdkit ≥ 1.1.12, and the default in nbdkit ≥ 1.3.

export names

Supported in nbdkit ≥ 1.1.12.

nbdkit can advertise an export name, but ignores any export name sent by the client. nbdkit does not support serving different data on different export names.

"NBD_FLAG_NO_ZEROES"

Supported in nbdkit ≥ 1.1.13.

This protocol optimization avoids sending a useless block of zero bytes during protocol negotiation.

"NBD_CMD_WRITE_ZEROES"

Supported in nbdkit ≥ 1.1.13.

TLS with X.509 certificates

Supported in nbdkit ≥ 1.1.15.

TLS with Pre-Shared Keys (PSK)

Supported in nbdkit ≥ 1.4.0.

"NBD_OPT_GO"

Supported in nbdkit ≥ 1.5.3.

This protocol enhancement allows the server to return errors when negotiating the export name.

"NBD_OPT_INFO"

Supported in nbdkit ≥ 1.9.3.

This protocol enhancement allows a client to inspect details about the export without actually connecting.

"NBD_FLAG_CAN_MULTI_CONN"

Supported in nbdkit ≥ 1.9.9.

Structured Replies

Supported in nbdkit ≥ 1.11.8.

However we don’t expose the capability to send structured replies to plugins yet, nor do we send human-readable error messages using this facility.

In nbdkit ≥ 1.13.9>, the command-line option --no-sr can be used to disable server support for structured replies, for testing client fallbacks.

Metadata Querying

Supported in nbdkit ≥ 1.11.8.

Block Status

Supported in nbdkit ≥ 1.11.10.

Only "base:allocation" (ie. querying which parts of an image are sparse) is supported.

Sparse reads (using "NBD_REPLY_TYPE_OFFSET_HOLE" are not directly supported, but a client can use block status to infer which portions of the export do not need to be read.

"NBD_FLAG_DF"

Supported in nbdkit ≥ 1.11.11.

This protocol extension allows a client to force an all-or-none read when structured replies are in effect. However, the flag is a no-op until we extend the plugin API to allow a fragmented read in the first place.

"NBD_CMD_CACHE"

Supported in nbdkit ≥ 1.13.4.

This protocol extension allows a client to inform the server about intent to access a portion of the export, to allow the server an opportunity to cache things appropriately.

"NBD_CMD_FLAG_FAST_ZERO"

Supported in nbdkit ≥ 1.15.0.

This protocol extension allows a server to advertise that it can rank all zero requests as fast or slow, at which point the client can make fast zero requests which fail immediately with "ENOTSUP" if the request is no faster than a counterpart write would be, while normal zero requests still benefit from compressed network traffic regardless of the time taken.

Resize Extension

Not supported.

AUTHORS

Eric Blake

Richard W.M. Jones

Pino Toscano

COPYRIGHT

LICENSE

Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
Neither the name of Red Hat nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission.

THIS SOFTWARE IS PROVIDED BY RED HAT AND CONTRIBUTORS ''AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL RED HAT OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.