nbdkit-plugin - Man Page

Plugins written in C have an ABI guarantee: a plugin compiled against an older version of nbdkit will still work correctly when loaded with a newer nbdkit.  We also try (but cannot guarantee) to support plugins compiled against a newer version of nbdkit when loaded with an older nbdkit, although the plugin may have reduced functionality if it depends on features only provided by newer nbdkit.

For plugins written in C, we also provide an API guarantee: a plugin written against an older header will still compile unmodified with a newer nbdkit.

The API guarantee does not always apply to plugins written in other (non-C) languages which may have to adapt to changes when recompiled against a newer nbdkit.

Plugins must choose which API version they want to use, by defining NBDKIT_API_VERSION before including <nbdkit-plugin.h> (or any other nbdkit header).

If omitted, the default version is 1 for backwards-compatibility with nbdkit v1.1.26 and earlier; however, it is recommended that new plugins be written to the maximum version (currently 2) as it enables more features and better interaction with nbdkit filters.

The rest of this document only covers the version 2 interface.  A newer nbdkit will always support plugins written in C which use any prior API version.

All plugins should start by including this header file (after optionally choosing an API version).

All plugins must define a thread model.  See "Threads" below for details.  It is generally safe to use:

 #define THREAD_MODEL NBDKIT_THREAD_MODEL_SERIALIZE_ALL_REQUESTS

All plugins must define and register one struct nbdkit_plugin, which contains the name of the plugin and pointers to callback functions, and use the NBDKIT_REGISTER_PLUGIN(plugin) macro:

 static struct nbdkit_plugin plugin = {
   .name              = "myplugin",
   .longname          = "My Plugin",
   .description       = "This is my great plugin for nbdkit",
   .open              = myplugin_open,
   .get_size          = myplugin_get_size,
   .pread             = myplugin_pread,
   .pwrite            = myplugin_pwrite,
   /* etc */
 };
 NBDKIT_REGISTER_PLUGIN(plugin)

The .name field is the name of the plugin.

The callbacks are described below (see "Callbacks").  Only .name, .open, .get_size and .pread are required.  All other callbacks can be omitted, although typical plugins need to use more.

Callbacks are called in the following order over the lifecycle of the plugin:

         ┌──────────────────┐
         │ load             │
         └─────────┬────────┘
                   │           configuration phase starts     ─┐
         ┌─────────┴────────┐                                  ┆
         │ config           │  config is called once per       ┆
         └─────────┬────────┘↺ key=value on the command line   ┆
         ┌─────────┴────────┐                                  ┆
         │ config_complete  │                                  ┆
         └─────────┬────────┘                                  ┆
         ┌─────────┴────────┐                                  ┆
         │ thread_model     │                                  ┆
         └─────────┬────────┘  configuration phase ends       ─┘
         ┌─────────┴────────┐
         │ get_ready        │
         └─────────┬────────┘
                   │           nbdkit forks into the background
         ┌─────────┴────────┐
         │ after_fork       │
         └─────────┬────────┘
                   │           nbdkit starts serving clients
                   │
        ┌──────────┴─────────────┬─ ─ ─ ─ ─ ─ ─ ─ ─
 ┌──────┴─────┐ client #1        │
 │ preconnect │                  │
 └──────┬─────┘                  │
 ┌──────┴─────┐                  │
 │list_exports│                  │
 └──────┬─────┘                  │
 ┌──────┴─────┐                  │
 │ open       │                  │
 └──────┬─────┘                  │
 ┌──────┴─────┐  NBD option      │
 │ can_write  │  negotiation     │
 └──────┬─────┘                  │
 ┌──────┴─────┐           ┌──────┴─────┐ client #2
 │ get_size   │           │ preconnect │
 └──────┬─────┘           └──────┬─────┘
 ┌──────┴─────┐ data
 │ pread      │ serving
 └──────┬─────┘↺                ...
 ┌──────┴─────┐
 │ pwrite     │
 └──────┬─────┘↺          ┌──────┴─────┐
 ┌──────┴─────┐           │ close      │
 │ close      │           └────────────┘
 └────────────┘

                   │           before nbdkit exits
                   │
         ┌─────────┴────────┐
         │ cleanup          │
         └─────────┬────────┘
         ┌─────────┴────────┐
         │ unload           │
         └──────────────────┘

Notes

nbdkit --dump-plugin

The order of calls when the user queries the plugin is slightly different: .config is called once for each parameter. .config_complete is not called.  .thread_model is called after .config.

.get_ready

This is the last chance to do any global preparation that is needed to serve connections.  In particular, error messages here will be visible to the user, but they might not be in .after_fork (see below).

Plugins should not create background threads here.  Use .after_fork instead.

.after_fork

.after_fork is called after the server has forked into the background and changed UID and directory.  If a plugin needs to create background threads (or uses an external library that creates threads) it should do so here, because background threads are invalidated by fork.

Because the server may have forked into the background, error messages and failures from .after_fork cannot be seen by the user unless they look through syslog.  An error in .after_fork can appear to the user as if nbdkit “just died”.  So in almost all cases it is better to use .get_ready instead of this callback, or to do as much preparation work as possible in .get_ready and only start background threads here.

The server doesn't always fork (eg. if the -f flag is used), but even so this callback will be called.  If you want to find out if the server forked between .get_ready and .after_fork use getpid(2).

.preconnect and .open

.preconnect is called when a TCP connection has been made to the server.  This happens early, before NBD or TLS negotiation.

.open is called when a new client has connected and finished the NBD handshake.  TLS negotiation (if used) has been completed successfully.

.can_write, .get_size and other option negotiation callbacks

These are called during option negotiation with the client, but before any data is served.  These callbacks may return different values across different .open calls, but within a single connection, they are called at most once and cached by nbdkit for that connection.

.pread, .pwrite and other data serving callbacks

After option negotiation has finished, these may be called to serve data.  Depending on the thread model chosen, they might be called in parallel from multiple threads.  The data serving callbacks include a flags argument; the results of the negotiation callbacks influence whether particular flags will ever be passed to a data callback.

.cleanup and .unload

The difference between these two methods is that .cleanup is called before any filter has been removed from memory with dlclose(3). When .unload is called, nbdkit is in the middle of calling dlclose(3).  Most plugins do not need to worry about this difference.

The following flags are defined by nbdkit, and used in various data serving callbacks as follows:

NBDKIT_FLAG_MAY_TRIM

This flag is used by the .zero callback; there is no way to disable this flag, although a plugin that does not support trims as a way to write zeroes may ignore the flag without violating expected semantics.

NBDKIT_FLAG_FUA

This flag represents Forced Unit Access semantics.  It is used by the .pwrite, .zero, and .trim callbacks to indicate that the plugin must not return a result until the action has landed in persistent storage.  This flag will not be sent to the plugin unless .can_fua is provided and returns NBDKIT_FUA_NATIVE.

The following defines are valid as successful return values for .can_fua:

NBDKIT_FUA_NONE

Forced Unit Access is not supported; the client must manually request a flush after writes have completed.  The NBDKIT_FLAG_FUA flag will not be passed to the plugin's write callbacks.

NBDKIT_FUA_EMULATE

The client may request Forced Unit Access, but it is implemented by emulation, where nbdkit calls .flush after a write operation; this is semantically correct, but may hurt performance as it tends to flush more data than just what the client requested.  The NBDKIT_FLAG_FUA flag will not be passed to the plugin's write callbacks.

NBDKIT_FUA_NATIVE

The client may request Forced Unit Access, which results in the NBDKIT_FLAG_FUA flag being passed to the plugin's write callbacks (.pwrite, .trim, and .zero).  When the flag is set, these callbacks must not return success until the client's request has landed in persistent storage.

The following defines are valid as successful return values for .can_cache:

NBDKIT_CACHE_NONE

The server does not advertise caching support, and rejects any client-requested caching. Any .cache callback is ignored.

NBDKIT_CACHE_EMULATE

The nbdkit server advertises cache support to the client, where the client may request that the server cache a region of the export to potentially speed up future read and/or write operations on that region. The nbdkit server implements the caching by calling .pread and ignoring the results. This option exists to ease the implementation of a common form of caching; any .cache callback is ignored.

NBDKIT_CACHE_NATIVE

The nbdkit server advertises cache support to the client, where the client may request that the server cache a region of the export to potentially speed up future read and/or write operations on that region. The nbdkit server calls the .cache callback to perform the caching; if that callback is missing, the client's cache request succeeds without doing anything.

Each nbdkit plugin must declare its maximum thread safety model by defining the THREAD_MODEL macro.  (This macro is used by NBDKIT_REGISTER_PLUGIN).  Additionally, a plugin may implement the .thread_model callback, called right after .config_complete to make a runtime decision on which thread model to use.  The nbdkit server chooses the most restrictive model between the plugin's THREAD_MODEL, the .thread_model if present, any restrictions requested by filters, and any limitations imposed by the operating system.

In nbdkit --dump-plugin PLUGIN output, the max_thread_model line matches the THREAD_MODEL macro, and the thread_model line matches what the system finally settled on after applying all restrictions.

The possible settings for THREAD_MODEL are defined below.

#define THREAD_MODEL NBDKIT_THREAD_MODEL_SERIALIZE_CONNECTIONS

Only a single handle can be open at any time, and all requests happen from one thread.

Note this means only one client can connect to the server at any time. If a second client tries to connect it will block waiting for the first client to close the connection.

#define THREAD_MODEL NBDKIT_THREAD_MODEL_SERIALIZE_ALL_REQUESTS

This is a safe default for most plugins.

Multiple handles can be open at the same time, but requests are serialized so that for the plugin as a whole only one open/read/write/close (etc) request will be in progress at any time.

This is a useful setting if the library you are using is not thread-safe.  However performance may not be good.

#define THREAD_MODEL NBDKIT_THREAD_MODEL_SERIALIZE_REQUESTS

Multiple handles can be open and multiple data requests can happen in parallel.  However only one request will happen per handle at a time (but requests on different handles might happen concurrently).

#define THREAD_MODEL NBDKIT_THREAD_MODEL_PARALLEL

Multiple handles can be open and multiple data requests can happen in parallel (even on the same handle).  The server may reorder replies, answering a later request before an earlier one.

All the libraries you use must be thread-safe and reentrant, and any code that creates a file descriptor should atomically set FD_CLOEXEC if you do not want it accidentally leaked to another thread's child process.  You may also need to provide mutexes for fields in your connection handle.

If none of the above thread models are suitable, use NBDKIT_THREAD_MODEL_PARALLEL and implement your own locking using pthread_mutex_t etc.

If there is an error in the plugin, the plugin should call nbdkit_error(3) to report an error message.  Then, the callback should return the appropriate error indication, eg. NULL or -1.

More information about these functions and nbdkit_verror(3) can be found in the separate manual page about them.

 const char *name;

This field (a string) is required, and must contain only ASCII alphanumeric characters or non-leading dashes, and be unique amongst all plugins.

 const char *version;

Plugins may optionally set a version string which is displayed in help and debugging output.  (See also "Version" below)

 const char *longname;

An optional free text name of the plugin.  This field is used in error messages.

 const char *description;

An optional multi-line description of the plugin.

 void load (void);

This is called once just after the plugin is loaded into memory.  You can use this to perform any global initialization needed by the plugin.

 void unload (void);

This may be called once just before the plugin is unloaded from memory.  Note that it's not guaranteed that .unload will always be called (eg. the server might be killed or segfault), so you should try to make the plugin as robust as possible by not requiring cleanup. See also "Shutdown" below.

 void dump_plugin (void);

This optional callback is called when the nbdkit plugin --dump-plugin command is used.  It should print any additional informative key=value fields to stdout as needed. Prefixing the keys with the name of the plugin will avoid conflicts.

 int config (const char *key, const char *value);

On the nbdkit command line, after the plugin filename, come an optional list of key=value arguments.  These are passed to the plugin through this callback when the plugin is first loaded and before any connections are accepted.

This callback may be called zero or more times.

Both key and value parameters will be non-NULL.  The strings are owned by nbdkit but will remain valid for the lifetime of the plugin, so the plugin does not need to copy them.

The key will be a non-empty string beginning with an ASCII alphabetic character (A-Z a-z).  The rest of the key must contain only ASCII alphanumeric plus period, underscore or dash characters (A-Z a-z 0-9 . _ -).  The value may be an arbitrary string, including an empty string.

The names of keys accepted by plugins is up to the plugin, but you should probably look at other plugins and follow the same conventions.

If the value is a relative path, then note that the server changes directory when it starts up.  See nbdkit_realpath(3).

If nbdkit_stdio_safe(3) returns 1, the value of the configuration parameter may be used to trigger reading additional data through stdin (such as a password or inline script).

If the .config callback is not provided by the plugin, and the user tries to specify any key=value arguments, then nbdkit will exit with an error.

If there is an error, .config should call nbdkit_error(3) with an error message and return -1.

 const char *magic_config_key;

This optional string can be used to set a "magic" key used when parsing plugin parameters.  It affects how "bare parameters" (those which do not contain an = character) are parsed on the command line.

If magic_config_key != NULL then any bare parameters are passed to the .config method as: config (magic_config_key, argv[i]);.

If magic_config_key is not set then we behave as in nbdkit < 1.7: If the first parameter on the command line is bare then it is passed to the .config method as: config ("script", value);. Any other bare parameters give errors.

 int config_complete (void);

This optional callback is called after all the configuration has been passed to the plugin.  It is a good place to do checks, for example that the user has passed the required parameters to the plugin.

If there is an error, .config_complete should call nbdkit_error(3) with an error message and return -1.

 const char *config_help;

This optional multi-line help message should summarize any key=value parameters that it takes.  It does not need to repeat what already appears in .description.

If the plugin doesn't take any config parameters you should probably omit this.

 int thread_model (void)

This optional callback is called after all the configuration has been passed to the plugin.  It can be used to force a stricter thread model based on configuration, compared to THREAD_MODEL.  See "Threads" above for details.  Attempts to request a looser (more parallel) model are silently ignored.

If there is an error, .thread_model should call nbdkit_error(3) with an error message and return -1.

 int get_ready (void);

This optional callback is called before the server starts serving.  It is called before the server forks or changes directory.  It is ordinarily the last chance to do any global preparation that is needed to serve connections.

If there is an error, .get_ready should call nbdkit_error(3) with an error message and return -1.

 int after_fork (void);

This optional callback is called before the server starts serving.  It is called after the server forks and changes directory.  If a plugin needs to create background threads (or uses an external library that creates threads) it should do so here, because background threads are killed by fork.  However you should try to do as little as possible here because error reporting is difficult.  See "Callback lifecycle" above.

If there is an error, .after_fork should call nbdkit_error(3) with an error message and return -1.

 void cleanup (void);

This optional callback is called after the server has closed all connections and is preparing to unload.  It is only reached in the same cases that the .after_fork callback was used, making it a good place to clean up any background threads.  However, it is not guaranteed that this callback will be reached, so you should try to make the plugin as robust as possible by not requiring cleanup.  See also "Shutdown" below.

 int preconnect (int readonly);

This optional callback is called when a TCP connection has been made to the server.  This happens early, before NBD or TLS negotiation.  If TLS authentication is required to access the server, then it has not been negotiated at this point.

For security reasons (to avoid denial of service attacks) this callback should be written to be as fast and take as few resources as possible.  If you use this callback, only use it to do basic access control, such as checking nbdkit_peer_name(3), nbdkit_peer_pid(3), nbdkit_peer_uid(3), nbdkit_peer_gid(3) against a list of permitted source addresses (see "Peer Name" and nbdkit-ip-filter(1)).  It may be better to do access control outside the server, for example using TCP wrappers or a firewall.

The readonly flag informs the plugin that the server was started with the -r flag on the command line.

Returning 0 will allow the connection to continue.  If there is an error or you want to deny the connection, call nbdkit_error(3) with an error message and return -1.

 int list_exports (int readonly, int is_tls,
                   struct nbdkit_exports *exports);

This optional callback is called if the client tries to list the exports served by the plugin (using NBD_OPT_LIST).  If the plugin does not supply this callback then the result of .default_export is advertised as the lone export.  The NBD protocol defines "" as the default export, so this is suitable for plugins which ignore the export name and always serve the same content.  See also "EXPORT NAME" below.

The readonly flag informs the plugin that the server was started with the -r flag on the command line, which is the same value passed to .preconnect and .open.  However, the NBD protocol does not yet have a way to let the client advertise an intent to be read-only even when the server allows writes, so this parameter may not be as useful as it appears.

The is_tls flag informs the plugin whether this listing was requested after the client has completed TLS negotiation.  When running the server in a mode that permits but does not require TLS, be careful that any exports listed when is_tls is false do not leak unintended information.

The exports parameter is an opaque object for collecting the list of exports.  Call nbdkit_add_export as needed to add specific exports to the list.

 int nbdkit_add_export (struct nbdkit_export *exports,
                        const char *name, const char *description);

The name must be a non-NULL, UTF-8 string between 0 and 4096 bytes in length.  Export names must be unique.  description is an optional description of the export which some clients can display but which is otherwise unused (if you don't want a description, you can pass this parameter as NULL).  The string(s) are copied into the exports list so you may free them immediately after calling this function.  nbdkit_add_export returns 0 on success or -1 on failure; on failure nbdkit_error(3) has already been called, with errno set to a suitable value.

There are also situations where a plugin may wish to duplicate the nbdkit default behavior of supplying an export list containing only the result of .default_export when .list_exports is missing; this is most common in a language binding where it is not known at compile time whether the language script will be providing an implementation for .list_exports, and is done by calling nbdkit_use_default_export.

 int nbdkit_use_default_export (struct nbdkit_export *exports);

nbdkit_use_default_export returns 0 on success or -1 on failure; on failure nbdkit_error(3) has already been called, with errno set to a suitable value.

The plugin may also leave the export list empty, by not calling either helper.  Once the plugin is happy with the list contents, returning 0 will send the list of exports back to the client.  If there is an error, .list_exports should call nbdkit_error(3) with an error message and return -1.

 const char *default_export (int readonly, int is_tls);

This optional callback is called if the client tries to connect to the default export "", where the plugin provides a UTF-8 string between 0 and 4096 bytes.  If the plugin does not supply this callback, the connection continues with the empty name; if the plugin returns a valid string, nbdkit behaves as if the client had passed that string instead of an empty name, and returns that name to clients that support it (see the NBD_INFO_NAME response to NBD_OPT_GO). Similarly, if the plugin does not supply a .list_exports callback, the result of this callback determines what export name to advertise to a client requesting an export list.

The readonly flag informs the plugin that the server was started with the -r flag on the command line, which is the same value passed to .preconnect and .open.  However, the NBD protocol does not yet have a way to let the client advertise an intent to be read-only even when the server allows writes, so this parameter may not be as useful as it appears.

The is_tls flag informs the plugin whether the canonical name for the default export is being requested after the client has completed TLS negotiation.  When running the server in a mode that permits but does not require TLS, be careful that a default export name does not leak unintended information.

If the plugin returns NULL or an invalid string (such as longer than 4096 bytes), the client is not permitted to connect to the default export.  However, this is not an error in the protocol, so it is not necessary to call nbdkit_error(3).

 void *open (int readonly);

This is called when a new client connects to the nbdkit server.  The callback should allocate a handle and return it.  This handle is passed back to other callbacks and could be freed in the .close callback.

Note that the handle is completely opaque to nbdkit, but it must not be NULL.  If you don't need to use a handle, return NBDKIT_HANDLE_NOT_NEEDED which is a static non-NULL pointer.

The readonly flag informs the plugin that the server was started with the -r flag on the command line which forces connections to be read-only.  Note that the plugin may additionally force the connection to be readonly (even if this flag is false) by returning false from the .can_write callback.  So if your plugin can only serve read-only, you can ignore this parameter.

If the plugin wants to differentiate the content it serves based on client input, then this is the spot to use nbdkit_export_name(3) to determine which export the client requested.  See also "Export Name and TLS" below.

This callback is called after the NBD handshake has completed; if the server requires TLS authentication, then that has occurred as well. But if the server is set up to have optional TLS authentication, you may check nbdkit_is_tls(3) to learn whether the client has completed TLS authentication.  When running the server in a mode that permits but does not require TLS, be careful that you do not allow unauthenticated clients to cause a denial of service against authentication.

If there is an error, .open should call nbdkit_error(3) with an error message and return NULL.

 void close (void *handle);

This is called when the client closes the connection.  It should clean up any per-connection resources.

Note there is no way in the NBD protocol to communicate close errors back to the client, for example if your plugin calls close(2) and you are checking for errors (as you should do).  Therefore the best you can do is to log the error on the server.  Well-behaved NBD clients should try to flush the connection before it is closed and check for errors, but obviously this is outside the scope of nbdkit.

 int64_t get_size (void *handle);

This is called during the option negotiation phase of the protocol to get the size (in bytes) of the block device being exported.

The returned size must be ≥ 0.  If there is an error, .get_size should call nbdkit_error(3) with an error message and return -1.

 const char *export_description (void *handle);

This is called during the option negotiation phase only if the client specifically requested an export description (see the NBD_INFO_DESCRIPTION response to NBD_OPT_GO).  Any description provided must be human-readable UTF-8, no longer than 4096 bytes.  Ideally, this description should match any description set during .list_exports, but that is not enforced.

If the plugin returns NULL or an invalid string (such as longer than 4096 bytes), or if this callback is omitted, no description is offered to the client.  As this is not an error in the protocol, it is not necessary to call nbdkit_error(3).  If the callback will not be returning a compile-time constant string, you may find nbdkit_strdup_intern(3) helpful for returning a value that avoids a memory leak.

 int block_size (void *handle, uint32_t *minimum,
                 uint32_t *preferred, uint32_t *maximum);

This is called during the option negotiation phase of the protocol to get the minimum, preferred and maximum block size (all in bytes) of the block device.  The client should obey these constraints by not making requests which are smaller than the minimum size or larger than the maximum size, and usually making requests of a multiple of the preferred size.  Furthermore requests should be aligned to at least the minimum block size, and usually the preferred block size.

"Preferred" block size in the NBD specification can be misinterpreted. It means the I/O size which does not have a penalty for read-modify-write.

Even if the plugin implements this callback, this does not mean that all client requests will obey the constraints.  A client could still ignore the constraints.  nbdkit passes all requests through to the plugin, because what the plugin does depends on the plugin's policy.  It might decide to serve the requests correctly anyway, or reject them with an error.  Plugins can avoid this complexity by using nbdkit-blocksize-policy-filter(1) which allows both setting/adjusting the constraints, and selecting an error policy.

The minimum block size must be ≥ 1.  The maximum block size must be ≤ 0xffff_ffff.  minimum ≤ preferred ≤ maximum.

As a special case, the plugin may return minimum == preferred == maximum == 0, meaning no information.

If this callback is not used, then the NBD protocol assumes by default minimum = 1, preferred = 4096.  (Maximum block size depends on various factors, see the NBD protocol specification, section "Block size constraints").

 int can_write (void *handle);

This is called during the option negotiation phase to find out if the handle supports writes.

If there is an error, .can_write should call nbdkit_error(3) with an error message and return -1.

This callback is not required.  If omitted, then we return true iff a .pwrite callback has been defined.

 int can_flush (void *handle);

This is called during the option negotiation phase to find out if the handle supports the flush-to-disk operation.

If there is an error, .can_flush should call nbdkit_error(3) with an error message and return -1.

This callback is not required.  If omitted, then we return true iff a .flush callback has been defined.

 int is_rotational (void *handle);

This is called during the option negotiation phase to find out if the backing disk is a rotational medium (like a traditional hard disk) or not (like an SSD).  If true, this may cause the client to reorder requests to make them more efficient for a slow rotating disk.

If there is an error, .is_rotational should call nbdkit_error(3) with an error message and return -1.

This callback is not required.  If omitted, then we return false.

 int can_trim (void *handle);

This is called during the option negotiation phase to find out if the plugin supports the trim/discard operation for punching holes in the backing storage.

If there is an error, .can_trim should call nbdkit_error(3) with an error message and return -1.

This callback is not required.  If omitted, then we return true iff a .trim callback has been defined.

 int can_zero (void *handle);

This is called during the option negotiation phase to find out if the plugin wants the .zero callback to be utilized.  Support for writing zeroes is still advertised to the client (unless the nbdkit-nozero-filter(1) is also used), so returning false merely serves as a way to avoid complicating the .zero callback to have to fail with ENOTSUP or EOPNOTSUPP on the connections where it will never be more efficient than using .pwrite up front.

If there is an error, .can_zero should call nbdkit_error(3) with an error message and return -1.

This callback is not required.  If omitted, then for a normal zero request, nbdkit always tries .zero first if it is present, and gracefully falls back to .pwrite if .zero was absent or failed with ENOTSUP or EOPNOTSUPP.

 int can_fast_zero (void *handle);

This is called during the option negotiation phase to find out if the plugin wants to advertise support for fast zero requests.  If this support is not advertised, a client cannot attempt fast zero requests, and has no way to tell if writing zeroes offers any speedups compared to using .pwrite (other than compressed network traffic).  If support is advertised, then .zero will have NBDKIT_FLAG_FAST_ZERO set when the client has requested a fast zero, in which case the plugin must fail with ENOTSUP or EOPNOTSUPP up front if the request would not offer any benefits over .pwrite.  Advertising support for fast zero requests does not require that writing zeroes be fast, only that the result (whether success or failure) is fast, so this should be advertised when feasible.

If there is an error, .can_fast_zero should call nbdkit_error(3) with an error message and return -1.

This callback is not required.  If omitted, then nbdkit returns true if .zero is absent or .can_zero returns false (in those cases, nbdkit fails all fast zero requests, as its fallback to .pwrite is not inherently faster), otherwise false (since it cannot be determined in advance if the plugin's .zero will properly honor the semantics of NBDKIT_FLAG_FAST_ZERO).

 int can_extents (void *handle);

This is called during the option negotiation phase to find out if the plugin supports detecting allocated (non-sparse) regions of the disk with the .extents callback.

If there is an error, .can_extents should call nbdkit_error(3) with an error message and return -1.

This callback is not required.  If omitted, then we return true iff a .extents callback has been defined.

 int can_fua (void *handle);

This is called during the option negotiation phase to find out if the plugin supports the Forced Unit Access (FUA) flag on write, zero, and trim requests.  If this returns NBDKIT_FUA_NONE, FUA support is not advertised to the client; if this returns NBDKIT_FUA_EMULATE, the .flush callback must work (even if .can_flush returns false), and FUA support is emulated by calling .flush after any write operation; if this returns NBDKIT_FUA_NATIVE, then the .pwrite, .zero, and .trim callbacks (if implemented) must handle the flag NBDKIT_FLAG_FUA, by not returning until that action has landed in persistent storage.

If there is an error, .can_fua should call nbdkit_error(3) with an error message and return -1.

This callback is not required unless a plugin wants to specifically handle FUA requests.  If omitted, nbdkit checks whether .flush exists, and behaves as if this function returns NBDKIT_FUA_NONE or NBDKIT_FUA_EMULATE as appropriate.

 int can_multi_conn (void *handle);

This is called during the option negotiation phase to find out if the plugin is prepared to handle multiple connections from a single client.  If the plugin sets this to true then a client may try to open multiple connections to the nbdkit server and spread requests across all connections to maximize parallelism.  If the plugin sets it to false (which is the default) then well-behaved clients should only open a single connection, although we cannot control what clients do in practice.

Specifically it means that either the plugin does not cache requests at all.  Or if it does cache them then the effects of a .flush request or setting NBDKIT_FLAG_FUA on a request must be visible across all connections to the plugin before the plugin replies to that request.

Properly working clients should send the same export name for each of these connections.

If you use Linux nbd-client(8) option -C num with num > 1 then Linux checks this flag and will refuse to connect if .can_multi_conn is false.

If there is an error, .can_multi_conn should call nbdkit_error(3) with an error message and return -1.

This callback is not required.  If omitted, then we return false.

 int can_cache (void *handle);

This is called during the option negotiation phase to find out if the plugin supports a cache or prefetch operation.

This can return:

NBDKIT_CACHE_NONE

Cache support is not advertised to the client.

NBDKIT_CACHE_EMULATE

Caching is emulated by the server calling .pread and discarding the result.

NBDKIT_CACHE_NATIVE

Cache support is advertised to the client.  The .cache callback will be called if it exists, otherwise all cache requests instantly succeed.

If there is an error, .can_cache should call nbdkit_error(3) with an error message and return -1.

This callback is not required.  If omitted, then we return NBDKIT_CACHE_NONE if the .cache callback is missing, or NBDKIT_CACHE_NATIVE if it is defined.

 int pread (void *handle, void *buf, uint32_t count, uint64_t offset,
            uint32_t flags);

During the data serving phase, nbdkit calls this callback to read data from the backing store.  count bytes starting at offset in the backing store should be read and copied into buf.  nbdkit takes care of all bounds- and sanity-checking, so the plugin does not need to worry about that.

The parameter flags exists in case of future NBD protocol extensions; at this time, it will be 0 on input.

The callback must read the whole count bytes if it can.  The NBD protocol doesn't allow partial reads (instead, these would be errors). If the whole count bytes was read, the callback should return 0 to indicate there was no error.

If there is an error (including a short read which couldn't be recovered from), .pread should call nbdkit_error(3) with an error message, and nbdkit_set_error(3) to record an appropriate error (unless errno is sufficient), then return -1.

 int pwrite (void *handle, const void *buf, uint32_t count, uint64_t offset,
             uint32_t flags);

During the data serving phase, nbdkit calls this callback to write data to the backing store.  count bytes starting at offset in the backing store should be written using the data in buf.  nbdkit takes care of all bounds- and sanity-checking, so the plugin does not need to worry about that.

This function will not be called if .can_write returned false.  The parameter flags may include NBDKIT_FLAG_FUA on input based on the result of .can_fua.

The callback must write the whole count bytes if it can.  The NBD protocol doesn't allow partial writes (instead, these would be errors).  If the whole count bytes was written successfully, the callback should return 0 to indicate there was no error.

If there is an error (including a short write which couldn't be recovered from), .pwrite should call nbdkit_error(3) with an error message, and nbdkit_set_error(3) to record an appropriate error (unless errno is sufficient), then return -1.

 int flush (void *handle, uint32_t flags);

During the data serving phase, this callback is used to fdatasync(2) the backing store, ie. to ensure it has been completely written to a permanent medium.  If that is not possible then you can omit this callback.

This function will not be called directly by the client if .can_flush returned false; however, it may still be called by nbdkit if .can_fua returned NBDKIT_FUA_EMULATE.  The parameter flags exists in case of future NBD protocol extensions; at this time, it will be 0 on input.

If there is an error, .flush should call nbdkit_error(3) with an error message, and nbdkit_set_error(3) to record an appropriate error (unless errno is sufficient), then return -1.

 int trim (void *handle, uint32_t count, uint64_t offset, uint32_t flags);

During the data serving phase, this callback is used to "punch holes" in the backing store.  If that is not possible then you can omit this callback.

This function will not be called if .can_trim returned false.  The parameter flags may include NBDKIT_FLAG_FUA on input based on the result of .can_fua.

If there is an error, .trim should call nbdkit_error(3) with an error message, and nbdkit_set_error(3) to record an appropriate error (unless errno is sufficient), then return -1.

 int zero (void *handle, uint32_t count, uint64_t offset, uint32_t flags);

During the data serving phase, this callback is used to write count bytes of zeroes at offset in the backing store.

This function will not be called if .can_zero returned false.  On input, the parameter flags may include NBDKIT_FLAG_MAY_TRIM unconditionally, NBDKIT_FLAG_FUA based on the result of .can_fua, and NBDKIT_FLAG_FAST_ZERO based on the result of .can_fast_zero.

If NBDKIT_FLAG_MAY_TRIM is requested, the operation can punch a hole instead of writing actual zero bytes, but only if subsequent reads from the hole read as zeroes.

If NBDKIT_FLAG_FAST_ZERO is requested, the plugin must decide up front if the implementation is likely to be faster than a corresponding .pwrite; if not, then it must immediately fail with ENOTSUP or EOPNOTSUPP (whether by nbdkit_set_error(3) or errno) and preferably without modifying the exported image.  It is acceptable to always fail a fast zero request (as a fast failure is better than attempting the write only to find out after the fact that it was not fast after all).  Note that on Linux, support for ioctl(BLKZEROOUT) is insufficient for determining whether a zero request to a block device will be fast (because the kernel will perform a slow fallback when needed).

The callback must write the whole count bytes if it can.  The NBD protocol doesn't allow partial writes (instead, these would be errors).  If the whole count bytes was written successfully, the callback should return 0 to indicate there was no error.

If there is an error, .zero should call nbdkit_error(3) with an error message, and nbdkit_set_error(3) to record an appropriate error (unless errno is sufficient), then return -1.

If this callback is omitted, or if it fails with ENOTSUP or EOPNOTSUPP (whether by nbdkit_set_error(3) or errno), then .pwrite will be used as an automatic fallback except when the client requested a fast zero.

 int extents (void *handle, uint32_t count, uint64_t offset,
              uint32_t flags, struct nbdkit_extents *extents);

During the data serving phase, this callback is used to detect allocated, sparse and zeroed regions of the disk.

This function will not be called if .can_extents returned false. nbdkit's default behaviour in this case is to treat the whole virtual disk as if it was allocated.  Also, this function will not be called by a client that does not request structured replies (the --no-sr option of nbdkit can be used to test behavior when .extents is unavailable to the client).

The callback should detect and return the list of extents overlapping the range [offset...offset+count-1].  The extents parameter points to an opaque object which the callback should fill in by calling nbdkit_add_extent.  See "Extents list" below.

If there is an error, .extents should call nbdkit_error(3) with an error message, and nbdkit_set_error(3) to record an appropriate error (unless errno is sufficient), then return -1.

Extents list

The plugin extents callback is passed an opaque pointer struct nbdkit_extents *extents.  This structure represents a list of filesystem extents describing which areas of the disk are allocated, which are sparse (“holes”), and, if supported, which are zeroes.

The extents callback should scan the disk starting at offset and call nbdkit_add_extent for each extent found.

Extents overlapping the range [offset...offset+count-1] should be returned if possible.  However nbdkit ignores extents < offset so the plugin may, if it is easier to implement, return all extent information for the whole disk.  The plugin may return extents beyond the end of the range.  It may also return extent information for less than the whole range, but it must return at least one extent overlapping offset.

The extents must be added in ascending order, and must be contiguous.

The flags parameter of the .extents callback may contain the flag NBDKIT_FLAG_REQ_ONE.  This means that the client is only requesting information about the extent overlapping offset.  The plugin may ignore this flag, or as an optimization it may return just a single extent for offset.

 int nbdkit_add_extent (struct nbdkit_extents *extents,
                        uint64_t offset, uint64_t length, uint32_t type);

Add an extent covering [offset...offset+length-1] of one of the following four types:

type = 0

A normal, allocated data extent.

type = NBDKIT_EXTENT_HOLE|NBDKIT_EXTENT_ZERO

An unallocated extent, a.k.a. a “hole”, which reads back as zeroes. This is the normal type of hole applicable to most disks.

type = NBDKIT_EXTENT_ZERO

An allocated extent which is known to contain only zeroes.

type = NBDKIT_EXTENT_HOLE

An unallocated extent (hole) which does not read back as zeroes.  Note this should only be used in specialized circumstances such as when writing a plugin for (or to emulate) certain SCSI drives which do not guarantee that trimmed blocks read back as zeroes.

nbdkit_add_extent returns 0 on success or -1 on failure.  On failure nbdkit_error(3) and/or nbdkit_set_error(3) has already been called.  errno will be set to a suitable value.

 int cache (void *handle, uint32_t count, uint64_t offset, uint32_t flags);

During the data serving phase, this callback is used to give the plugin a hint that the client intends to make further accesses to the given region of the export.

The nature of caching/prefetching is not specified further by the NBD specification.  For example, a server may place limits on how much may be cached at once, and there is no way to control if writes to a cached area have write-through or write-back semantics.  In fact, the cache command can always fail and still be compliant, and success might not guarantee a performance gain.

If this callback is omitted, then the results of .can_cache determine whether nbdkit will reject cache requests, treat them as instant success, or emulate caching by calling .pread over the same region and ignoring the results.

This function will not be called if .can_cache did not return NBDKIT_CACHE_NATIVE.

The flags parameter exists in case of future NBD protocol extensions; at this time, it will be 0 on input.  A plugin must fail this function if flags includes an unrecognized flag, as that may indicate a requirement that the plugin must comply with to provide a specific caching semantic.

If there is an error, .cache should call nbdkit_error(3) with an error message, and nbdkit_set_error(3) to record an appropriate error (unless errno is sufficient), then return -1.

This field defaults to 0; if non-zero, nbdkit can reliably use the value of errno when a callback reports failure, rather than the plugin having to call nbdkit_set_error(3).

The macros NBDKIT_VERSION_MAJOR, NBDKIT_VERSION_MINOR and NBDKIT_VERSION_MICRO expand to integers containing the version of the nbdkit headers that you are compiling against.

NBDKIT_VERSION_MAJOR is always 1.  NBDKIT_VERSION_MINOR is even for stable releases of nbdkit and odd for development releases.

The macro NBDKIT_VERSION_STRING expands to the same version as a string.

When the plugin is loaded into nbdkit, it may not be the same version that it was compiled against.  nbdkit guarantees backwards compatibility of the API and ABI, so provided that nbdkit is the same version or newer, the plugin will still work.  There is no way to get the version of nbdkit from the plugin.

The plugin itself can use any versioning scheme you want, and put any string into the .version field of the plugin struct (or leave the field NULL).

See "Writing an Nbdkit Plugin" above.

The -v option switches general debugging on or off, and this debugging should be used for messages which are useful for all users of your plugin.

In cases where you want to enable specific extra debugging to track down bugs in plugins or filters — mainly for use by the plugin/filter developers themselves — you can define Debug Flags.  These are global ints called myplugin_debug_*:

 int myplugin_debug_foo;
 int myplugin_debug_bar;
 ...
 if (myplugin_debug_foo) {
   nbdkit_debug ("lots of extra debugging about foo: ...");
 }

Debug Flags can be controlled on the command line using the -D (or --debug) option:

 nbdkit -f -v -D myplugin.foo=1 -D myplugin.bar=2 myplugin [...]

Note myplugin is the name passed to .name in the struct nbdkit_plugin.

You should only use this feature for debug settings.  For general settings use ordinary plugin parameters.  Debug Flags can only be C ints.  They are not supported by non-C language plugins.

For convenience '.' characters are replaced with '_' characters in the variable name, so both of these parameters:

 -D myplugin.foo_bar=1
 -D myplugin.foo.bar=1

correspond to the plugin variable myplugin_debug_foo_bar.

nbdkit provides a pkg-config/pkgconf file called nbdkit.pc which should be installed on the correct path when the nbdkit plugin development environment is installed.  You can use this in autoconf configure.ac scripts to test for the development environment:

 PKG_CHECK_MODULES([NBDKIT], [nbdkit >= 1.2.3])

The above will fail unless nbdkit ≥ 1.2.3 and the header file is installed, and will set NBDKIT_CFLAGS and NBDKIT_LIBS appropriately for compiling plugins.

You can also run pkg-config/pkgconf directly, for example:

 if ! pkg-config nbdkit --exists; then
   echo "you must install the nbdkit plugin development environment"
   exit 1
 fi

You can also substitute the plugindir variable by doing:

 PKG_CHECK_VAR([NBDKIT_PLUGINDIR], [nbdkit], [plugindir])

which defines $(NBDKIT_PLUGINDIR) in automake-generated Makefiles.

If nbdkit development headers are installed in a non-standard location then you may need to compile plugins using:

 gcc -fPIC -shared myplugin.c -o myplugin.so \
   `pkg-config nbdkit --cflags --libs`