Server implementation

ServerInterface is an interface to override for server support.

class paramiko.server.InteractiveQuery(name='', instructions='', *prompts)

A query (set of prompts) for a user during interactive authentication.

__init__(name='', instructions='', *prompts)

Create a new interactive query to send to the client. The name and instructions are optional, but are generally displayed to the end user. A list of prompts may be included, or they may be added via the add_prompt method.

Parameters:
  • name (str) – name of this query
  • instructions (str) – user instructions (usually short) about this query
  • prompts (str) – one or more authentication prompts
__weakref__

list of weak references to the object (if defined)

add_prompt(prompt, echo=True)

Add a prompt to this query. The prompt should be a (reasonably short) string. Multiple prompts can be added to the same query.

Parameters:
  • prompt (str) – the user prompt
  • echo (bool) – True (default) if the user’s response should be echoed; False if not (for a password or similar)
class paramiko.server.ServerInterface

This class defines an interface for controlling the behavior of Paramiko in server mode.

Methods on this class are called from Paramiko’s primary thread, so you shouldn’t do too much work in them. (Certainly nothing that blocks or sleeps.)

__weakref__

list of weak references to the object (if defined)

cancel_port_forward_request(address, port)

The client would like to cancel a previous port-forwarding request. If the given address and port is being forwarded across this ssh connection, the port should be closed.

Parameters:
  • address (str) – the forwarded address
  • port (int) – the forwarded port
check_auth_gssapi_keyex(username, gss_authenticated=2, cc_file=None)

Authenticate the given user to the server if he is a valid krb5 principal and GSS-API Key Exchange was performed. If GSS-API Key Exchange was not performed, this authentication method won’t be available.

Parameters:
  • username (str) – The username of the authenticating client
  • gss_authenticated (int) – The result of the krb5 authentication
  • cc_filename (str) – The krb5 client credentials cache filename
Returns:

AUTH_FAILED if the user is not authenticated otherwise AUTH_SUCCESSFUL

Return type:

int

Note:

Kerberos credential delegation is not supported.

See:

ssh_gss kex_gss

Note:

: We are just checking in L{AuthHandler} that the given user is a valid krb5 principal! We don’t check if the krb5 principal is allowed to log in on the server, because there is no way to do that in python. So if you develop your own SSH server with paramiko for a cetain plattform like Linux, you should call C{krb5_kuserok()} in your local kerberos library to make sure that the krb5_principal has an account on the server and is allowed to log in as a user.

See:

http://www.unix.com/man-page/all/3/krb5_kuserok/

check_auth_gssapi_with_mic(username, gss_authenticated=2, cc_file=None)

Authenticate the given user to the server if he is a valid krb5 principal.

Parameters:
  • username (str) – The username of the authenticating client
  • gss_authenticated (int) – The result of the krb5 authentication
  • cc_filename (str) – The krb5 client credentials cache filename
Returns:

AUTH_FAILED if the user is not authenticated otherwise AUTH_SUCCESSFUL

Return type:

int

Note:

Kerberos credential delegation is not supported.

See:

ssh_gss

Note:

: We are just checking in L{AuthHandler} that the given user is a valid krb5 principal! We don’t check if the krb5 principal is allowed to log in on the server, because there is no way to do that in python. So if you develop your own SSH server with paramiko for a cetain plattform like Linux, you should call C{krb5_kuserok()} in your local kerberos library to make sure that the krb5_principal has an account on the server and is allowed to log in as a user.

See:

http://www.unix.com/man-page/all/3/krb5_kuserok/

check_auth_interactive(username, submethods)

Begin an interactive authentication challenge, if supported. You should override this method in server mode if you want to support the "keyboard-interactive" auth type, which requires you to send a series of questions for the client to answer.

Return AUTH_FAILED if this auth method isn’t supported. Otherwise, you should return an InteractiveQuery object containing the prompts and instructions for the user. The response will be sent via a call to check_auth_interactive_response.

The default implementation always returns AUTH_FAILED.

Parameters:
  • username (str) – the username of the authenticating client
  • submethods (str) – a comma-separated list of methods preferred by the client (usually empty)
Returns:

AUTH_FAILED if this auth method isn’t supported; otherwise an object containing queries for the user

Return type:

int or InteractiveQuery

check_auth_interactive_response(responses)

Continue or finish an interactive authentication challenge, if supported. You should override this method in server mode if you want to support the "keyboard-interactive" auth type.

Return AUTH_FAILED if the responses are not accepted, AUTH_SUCCESSFUL if the responses are accepted and complete the authentication, or AUTH_PARTIALLY_SUCCESSFUL if your authentication is stateful, and this set of responses is accepted for authentication, but more authentication is required. (In this latter case, get_allowed_auths will be called to report to the client what options it has for continuing the authentication.)

If you wish to continue interactive authentication with more questions, you may return an InteractiveQuery object, which should cause the client to respond with more answers, calling this method again. This cycle can continue indefinitely.

The default implementation always returns AUTH_FAILED.

Parameters:responses (list) – list of str responses from the client
Returns:AUTH_FAILED if the authentication fails; AUTH_SUCCESSFUL if it succeeds; AUTH_PARTIALLY_SUCCESSFUL if the interactive auth is successful, but authentication must continue; otherwise an object containing queries for the user
Return type:int or InteractiveQuery
check_auth_none(username)

Determine if a client may open channels with no (further) authentication.

Return AUTH_FAILED if the client must authenticate, or AUTH_SUCCESSFUL if it’s okay for the client to not authenticate.

The default implementation always returns AUTH_FAILED.

Parameters:username (str) – the username of the client.
Returns:AUTH_FAILED if the authentication fails; AUTH_SUCCESSFUL if it succeeds.
Return type:int
check_auth_password(username, password)

Determine if a given username and password supplied by the client is acceptable for use in authentication.

Return AUTH_FAILED if the password is not accepted, AUTH_SUCCESSFUL if the password is accepted and completes the authentication, or AUTH_PARTIALLY_SUCCESSFUL if your authentication is stateful, and this key is accepted for authentication, but more authentication is required. (In this latter case, get_allowed_auths will be called to report to the client what options it has for continuing the authentication.)

The default implementation always returns AUTH_FAILED.

Parameters:
  • username (str) – the username of the authenticating client.
  • password (str) – the password given by the client.
Returns:

AUTH_FAILED if the authentication fails; AUTH_SUCCESSFUL if it succeeds; AUTH_PARTIALLY_SUCCESSFUL if the password auth is successful, but authentication must continue.

Return type:

int

check_auth_publickey(username, key)

Determine if a given key supplied by the client is acceptable for use in authentication. You should override this method in server mode to check the username and key and decide if you would accept a signature made using this key.

Return AUTH_FAILED if the key is not accepted, AUTH_SUCCESSFUL if the key is accepted and completes the authentication, or AUTH_PARTIALLY_SUCCESSFUL if your authentication is stateful, and this password is accepted for authentication, but more authentication is required. (In this latter case, get_allowed_auths will be called to report to the client what options it has for continuing the authentication.)

Note that you don’t have to actually verify any key signtature here. If you’re willing to accept the key, Paramiko will do the work of verifying the client’s signature.

The default implementation always returns AUTH_FAILED.

Parameters:
  • username (str) – the username of the authenticating client
  • key (PKey) – the key object provided by the client
Returns:

AUTH_FAILED if the client can’t authenticate with this key; AUTH_SUCCESSFUL if it can; AUTH_PARTIALLY_SUCCESSFUL if it can authenticate with this key but must continue with authentication

Return type:

int

check_channel_direct_tcpip_request(chanid, origin, destination)

Determine if a local port forwarding channel will be granted, and return OPEN_SUCCEEDED or an error code. This method is called in server mode when the client requests a channel, after authentication is complete.

The chanid parameter is a small number that uniquely identifies the channel within a Transport. A Channel object is not created unless this method returns OPEN_SUCCEEDED – once a Channel object is created, you can call Channel.get_id to retrieve the channel ID.

The origin and destination parameters are (ip_address, port) tuples that correspond to both ends of the TCP connection in the forwarding tunnel.

The return value should either be OPEN_SUCCEEDED (or 0) to allow the channel request, or one of the following error codes to reject it:

  • OPEN_FAILED_ADMINISTRATIVELY_PROHIBITED
  • OPEN_FAILED_CONNECT_FAILED
  • OPEN_FAILED_UNKNOWN_CHANNEL_TYPE
  • OPEN_FAILED_RESOURCE_SHORTAGE

The default implementation always returns OPEN_FAILED_ADMINISTRATIVELY_PROHIBITED.

Parameters:
  • chanid (int) – ID of the channel
  • origin (tuple) – 2-tuple containing the IP address and port of the originator (client side)
  • destination (tuple) – 2-tuple containing the IP address and port of the destination (server side)
Returns:

an int success or failure code (listed above)

check_channel_env_request(channel, name, value)

Check whether a given environment variable can be specified for the given channel. This method should return True if the server is willing to set the specified environment variable. Note that some environment variables (e.g., PATH) can be exceedingly dangerous, so blindly allowing the client to set the environment is almost certainly not a good idea.

The default implementation always returns False.

Parameters:
  • channel – the Channel the env request arrived on
  • name (str) – name
  • value (str) – Channel value
Returns:

A boolean

check_channel_exec_request(channel, command)

Determine if a shell command will be executed for the client. If this method returns True, the channel should be connected to the stdin, stdout, and stderr of the shell command.

The default implementation always returns False.

Parameters:
  • channel (Channel) – the Channel the request arrived on.
  • command (str) – the command to execute.
Returns:

True if this channel is now hooked up to the stdin, stdout, and stderr of the executing command; False if the command will not be executed.

New in version 1.1.

check_channel_forward_agent_request(channel)

Determine if the client will be provided with an forward agent session. If this method returns True, the server will allow SSH Agent forwarding.

The default implementation always returns False.

Parameters:channel (Channel) – the Channel the request arrived on
Returns:True if the AgentForward was loaded; False if not
check_channel_pty_request(channel, term, width, height, pixelwidth, pixelheight, modes)

Determine if a pseudo-terminal of the given dimensions (usually requested for shell access) can be provided on the given channel.

The default implementation always returns False.

Parameters:
  • channel (Channel) – the Channel the pty request arrived on.
  • term (str) – type of terminal requested (for example, "vt100").
  • width (int) – width of screen in characters.
  • height (int) – height of screen in characters.
  • pixelwidth (int) – width of screen in pixels, if known (may be 0 if unknown).
  • pixelheight (int) – height of screen in pixels, if known (may be 0 if unknown).
Returns:

True if the pseudo-terminal has been allocated; False otherwise.

check_channel_request(kind, chanid)

Determine if a channel request of a given type will be granted, and return OPEN_SUCCEEDED or an error code. This method is called in server mode when the client requests a channel, after authentication is complete.

If you allow channel requests (and an ssh server that didn’t would be useless), you should also override some of the channel request methods below, which are used to determine which services will be allowed on a given channel:

The chanid parameter is a small number that uniquely identifies the channel within a Transport. A Channel object is not created unless this method returns OPEN_SUCCEEDED – once a Channel object is created, you can call Channel.get_id to retrieve the channel ID.

The return value should either be OPEN_SUCCEEDED (or 0) to allow the channel request, or one of the following error codes to reject it:

  • OPEN_FAILED_ADMINISTRATIVELY_PROHIBITED
  • OPEN_FAILED_CONNECT_FAILED
  • OPEN_FAILED_UNKNOWN_CHANNEL_TYPE
  • OPEN_FAILED_RESOURCE_SHORTAGE

The default implementation always returns OPEN_FAILED_ADMINISTRATIVELY_PROHIBITED.

Parameters:
  • kind (str) – the kind of channel the client would like to open (usually "session").
  • chanid (int) – ID of the channel
Returns:

an int success or failure code (listed above)

check_channel_shell_request(channel)

Determine if a shell will be provided to the client on the given channel. If this method returns True, the channel should be connected to the stdin/stdout of a shell (or something that acts like a shell).

The default implementation always returns False.

Parameters:channel (Channel) – the Channel the request arrived on.
Returns:True if this channel is now hooked up to a shell; False if a shell can’t or won’t be provided.
check_channel_subsystem_request(channel, name)

Determine if a requested subsystem will be provided to the client on the given channel. If this method returns True, all future I/O through this channel will be assumed to be connected to the requested subsystem. An example of a subsystem is sftp.

The default implementation checks for a subsystem handler assigned via Transport.set_subsystem_handler. If one has been set, the handler is invoked and this method returns True. Otherwise it returns False.

Note

Because the default implementation uses the Transport to identify valid subsystems, you probably won’t need to override this method.

Parameters:
  • channel (Channel) – the Channel the pty request arrived on.
  • name (str) – name of the requested subsystem.
Returns:

True if this channel is now hooked up to the requested subsystem; False if that subsystem can’t or won’t be provided.

check_channel_window_change_request(channel, width, height, pixelwidth, pixelheight)

Determine if the pseudo-terminal on the given channel can be resized. This only makes sense if a pty was previously allocated on it.

The default implementation always returns False.

Parameters:
  • channel (Channel) – the Channel the pty request arrived on.
  • width (int) – width of screen in characters.
  • height (int) – height of screen in characters.
  • pixelwidth (int) – width of screen in pixels, if known (may be 0 if unknown).
  • pixelheight (int) – height of screen in pixels, if known (may be 0 if unknown).
Returns:

True if the terminal was resized; False if not.

check_channel_x11_request(channel, single_connection, auth_protocol, auth_cookie, screen_number)

Determine if the client will be provided with an X11 session. If this method returns True, X11 applications should be routed through new SSH channels, using Transport.open_x11_channel.

The default implementation always returns False.

Parameters:
  • channel (Channel) – the Channel the X11 request arrived on
  • single_connection (bool) – True if only a single X11 channel should be opened, else False.
  • auth_protocol (str) – the protocol used for X11 authentication
  • auth_cookie (str) – the cookie used to authenticate to X11
  • screen_number (int) – the number of the X11 screen to connect to
Returns:

True if the X11 session was opened; False if not

check_global_request(kind, msg)

Handle a global request of the given kind. This method is called in server mode and client mode, whenever the remote host makes a global request. If there are any arguments to the request, they will be in msg.

There aren’t any useful global requests defined, aside from port forwarding, so usually this type of request is an extension to the protocol.

If the request was successful and you would like to return contextual data to the remote host, return a tuple. Items in the tuple will be sent back with the successful result. (Note that the items in the tuple can only be strings, ints, longs, or bools.)

The default implementation always returns False, indicating that it does not support any global requests.

Note

Port forwarding requests are handled separately, in check_port_forward_request.

Parameters:
  • kind (str) – the kind of global request being made.
  • msg (Message) – any extra arguments to the request.
Returns:

True or a tuple of data if the request was granted; False otherwise.

check_port_forward_request(address, port)

Handle a request for port forwarding. The client is asking that connections to the given address and port be forwarded back across this ssh connection. An address of "0.0.0.0" indicates a global address (any address associated with this server) and a port of 0 indicates that no specific port is requested (usually the OS will pick a port).

The default implementation always returns False, rejecting the port forwarding request. If the request is accepted, you should return the port opened for listening.

Parameters:
  • address (str) – the requested address
  • port (int) – the requested port
Returns:

the port number (int) that was opened for listening, or False to reject

enable_auth_gssapi()

Overwrite this function in your SSH server to enable GSSAPI authentication. The default implementation always returns false.

Returns bool:Whether GSSAPI authentication is enabled.
See:ssh_gss
get_allowed_auths(username)

Return a list of authentication methods supported by the server. This list is sent to clients attempting to authenticate, to inform them of authentication methods that might be successful.

The “list” is actually a string of comma-separated names of types of authentication. Possible values are "password", "publickey", and "none".

The default implementation always returns "password".

Parameters:username (str) – the username requesting authentication.
Returns:a comma-separated str of authentication types
get_banner()

A pre-login banner to display to the user. The message may span multiple lines separated by crlf pairs. The language should be in rfc3066 style, for example: en-US

The default implementation always returns (None, None).

Returns:A tuple containing the banner and language code.

New in version 2.3.

class paramiko.server.SubsystemHandler(channel, name, server)

Handler for a subsytem in server mode. If you create a subclass of this class and pass it to Transport.set_subsystem_handler, an object of this class will be created for each request for this subsystem. Each new object will be executed within its own new thread by calling start_subsystem. When that method completes, the channel is closed.

For example, if you made a subclass MP3Handler and registered it as the handler for subsystem "mp3", then whenever a client has successfully authenticated and requests subsytem "mp3", an object of class MP3Handler will be created, and start_subsystem will be called on it from a new thread.

__init__(channel, name, server)

Create a new handler for a channel. This is used by ServerInterface to start up a new handler when a channel requests this subsystem. You don’t need to override this method, but if you do, be sure to pass the channel and name parameters through to the original __init__ method here.

Parameters:
  • channel (Channel) – the channel associated with this subsystem request.
  • name (str) – name of the requested subsystem.
  • server (ServerInterface) – the server object for the session that started this subsystem
finish_subsystem()

Perform any cleanup at the end of a subsystem. The default implementation just closes the channel.

New in version 1.1.

get_server()

Return the ServerInterface object associated with this channel and subsystem.

start_subsystem(name, transport, channel)

Process an ssh subsystem in server mode. This method is called on a new object (and in a new thread) for each subsystem request. It is assumed that all subsystem logic will take place here, and when the subsystem is finished, this method will return. After this method returns, the channel is closed.

The combination of transport and channel are unique; this handler corresponds to exactly one Channel on one Transport.

Note

It is the responsibility of this method to exit if the underlying Transport is closed. This can be done by checking Transport.is_active or noticing an EOF on the Channel. If this method loops forever without checking for this case, your Python interpreter may refuse to exit because this thread will still be running.

Parameters:
  • name (str) – name of the requested subsystem.
  • transport (Transport) – the server-mode Transport.
  • channel (Channel) – the channel associated with this subsystem request.