pypsrp is a Python client for the PowerShell Remoting Protocol (PSRP) and Windows Remote Management (WinRM) service. It allows your to execute commands on a remote Windows host from any machine that can run Python.
This library exposes 4 different types of APIs;
- A simple client API that can copy files to and from the remote Windows host as well as execute processes and PowerShell scripts
- A WSMan interface to execute various WSMan calls like
Send
,Create
,Connect
,Disconnect
, etc - A Windows Remote Shell (WinRS) layer that executes cmd commands and executables using the base WinRM protocol
- A PowerShell Remoting Protocol (PSRP) layer allows you to create remote Runspace Pools and PowerShell pipelines
At a basic level, you can use this library to;
- Execute a cmd command
- Run another executable
- Execute PowerShell scripts
- Copy a file from the localhost to the remote Windows host
- Fetch a file from the remote Windows host to the localhost
- Create a Runspace Pool that contains one or multiple PowerShell pipelines and execute them asynchronously
- Support for a reference host base implementation of PSRP for interactive scripts
Currently this library only supports the WSMan transport method but is designed to support SSH at some point in the future (PR's are welcome). By default it supports the following authentication methods with WSMan;
- Basic
- Certificate
- NTLM
It also supports Negotiate/Kerberos
, and CredSSP
but require extra
libraries to be isntalled.
See How to Install
for more details
- CPython 2.6+
- cryptography
- pyspnego
- requests
The following Python libraries can be installed to add extra features that do not come with the base package
- python-gssapi for Kerberos authentication on Linux
- requests-credssp for CredSSP authentication
To install pypsrp with all basic features, run
pip install pypsrp
While pypsrp supports Kerberos authentication, it isn't included by default for Linux hosts due to it's reliance on system packages to be present.
To install these packages, run the below
For Debian/Ubuntu
# For Python 2
apt-get install gcc python-dev libkrb5-dev
# For Python 3
apt-get install gcc python3-dev libkrb5-dev
# To add NTLM to the GSSAPI SPNEGO auth run
apt-get install gss-ntlmssp
For RHEL/Centos
yum install gcc python-devel krb5-devel
# To add NTLM to the GSSAPI SPNEGO auth run
yum install gssntlmssp
For Fedora
dnf install gcc python-devel krb5-devel
# To add NTLM to the GSSAPI SPNEGO auth run
dnf install gssntlmssp
For Arch Linux
pacman -S gcc krb5
Once installed you can install the Python packages with
pip install pypsrp[kerberos]
Kerberos also needs to be configured to talk to the domain but that is outside the scope of this page.
Like Kerberos auth, CredSSP is supported but isn't included by default. To add support for CredSSP auth try to run the following
pip install pypsrp[credssp]
If that fails you may need to update pip and setuptools to a newer version
pip install -U pip setuptools
, otherwise the following system package may be
required;
# For Debian/Ubuntu
apt-get install gcc python-dev
# For RHEL/Centos
yum install gcc python-devel
# For Fedora
dnf install gcc python-devel
There are 3 main components that are in use within this library;
Transport
: Handles the raw transport of messages to and from the serverShell
: Handles the WSMV or PSRP protocol details used to create the remote shell that processes are run on, usesConnection
to send the detailsProcess
: Runs the process or script within a shell
Currently only the connection that is supported is the WSMan protocol over HTTP
through pypsrp.wsman.WSMan
and offers mostly all the same features in the
WSMV spec including;
- Basic, Certificate, Negotiate, Kerberos, and CredSSP authentication
- TLS encryption
- Message encryption with Negotiate, Kerberos, and CredSSP authentication
- Definable proxy
These are the options that can be used to setup WSMan
;
server
: The hostname or IP address of the host to connect tomax_envelope_size
: The maximum envelope size, in bytes, that can be sent to the server, default is153600
operation_timeout
: The operation timeout, in seconds, of each WSMan operation, default is20
. This should always be lower thanread_timeout
.port
: The port to connect to, default is5986
ifssl=True
else5985
username
: The username to connect with, required for all auths exceptcertificate
and optionally required fornegotiate/kerberos
password
: The password forusername
. Due to a bug on MacOS/Heimdal GSSAPI implementations, this will persist in the user's ccache when using Negotiate or Kerberos authentication, runkdestroy
manually to remove thisssl
: Whether to connect overhttps
orhttps
, default isTrue
path
: The WinRM path to connect to, default iswsman
auth
: The authentication protocol to use, default isnegotiate
, choices arebasic
,certificate
,negotiate
,ntlm
,kerberos
,credssp
cert_validation
: Whether to validate the server's SSL certificate, default isTrue
. Can beFalse
to not validate or a path to a PEM file of trusted certificatesconnection_timeout
: The timeout for creating a HTTP connection, default is30
read_timeout
: The timeout for receiving a response from the server after a request has been made, default is30
encryption
: Controls the encryption settings, default isauto
, choices areauto
,always
,never
. Set toalways
to always run message encryption even over HTTPS,never
to never use message encryption even over HTTPproxy
: The proxy URL used to connect to the remote hostno_proxy
: Whether to ignore any environment proxy variable and connect directly to the host, default isFalse
locale
: Thewsmv:Locale
value to set on each WSMan request. This specifies the language in which the cleint wants response text to be translated, default isen-US
data_locale
: Thewsmv:DataLocale
value to set on each WSMan request. This specifies the format in which numerical data is presented in the response text, default is the value oflocale
reconnection_retries
: Number of retries on a connection problem, default is0
reconnection_backoff
: Number of seconds to backoff in between reconnection attempts (first sleeps X, then sleeps 2X, 4X, 8*X, ...), default is2.0
certificate_key_pem
: The path to the certificate key used incertificate
authenticationcertificate_pem
: The path to the certificate used incertificate
authenticationcredssp_auth_mechanism
: The sub-auth mechanism used in CredSSP, default isauto
, choices areauto
,ntlm
, orkerberos
credssp_disable_tlsv1_2
: Whether to used CredSSP auth over the insecure TLSv1.0, default isFalse
credssp_minimum_version
: The minimum CredSSP server version that the client will connect to, default is2
negotiate_delegate
: Whether to negotiate the credential to the host, default isFalse
. This is only valid ifnegotiate
auth negotiated Kerberos orkerberos
was explicitly setnegotiate_hostname_override
: The hostname used to calculate the host SPN when authenticating the host with Kerberos auth. This is only valid ifnegotiate
auth negotiated Kerberos orkerberos
was explicitly setnegotiate_send_cbt
: Whether to binding the channel binding token (HTTPS only) to the auth or ignore, default isTrue
negotiate_service
: Override the service part of the calculated SPN used when authenticating the server, default isWSMAN
. This is only valid ifnegotiate
auth negotiated Kerberos orkerberos
was explicitly set
When running over HTTP, this library will enforce encryption by default but if
that is not supported (Basic auth) or isn't available on the host then either
use HTTPS or disable encryption with encryption="never"
.
There are plans to add support for SSH as a connection but this still needs to be implemented. SSH will work on hosts that are running PowerShell Core but not the standard PowerShell.
There are two shells that can be used in this library, pypsrp.shell.WinRS
and
pypsrp.powershell.RunspacePool
.
WinRS
is a cmd shell that can be used to issue cmd commands, including but
not limited to other executables. Here are the options that can be used to
configure a WinRS
shell;
wsman
: WinRS only works over WSMan, so this is thepypsrp.wsman.WSMan
object to run the commands overresource_uri
: The resource uri of the shell, defaults tohttp://schemas.microsoft.com/wbem/wsman/1/windows/shell/cmd
id
: The ID if the shell, this should be kept asNone
as it is created dynamically by the serverinput_streams
: The input stream(s) of the shell, default isstdin
output_streams
: The output stream(s) of the shell, default isstdout, stderr
codepage
: The codepage of the shell, default is the default of the hostenvironment
: A dictionary of environment key/values to set for the remote shellidle_time_out
: THe idle timeout in seconds of the shelllifetime
: The total lifetime of the shellname
: The name (description only) of the shellno_profile
: Whether to create the shell with the user profile loaded or notworking_directory
: The default working directory of the created shell
RunspacePool
is a shell used by the PSRP protocol, it is designed to be a
close implementation of the .NET
System.Management.Automation.Runspaces.RunspacePool
class. The methods and properties are similar and can mostly do the same thing.
Here are the options that can be used to configure a RunspacePool
shell;
connection
: The connection object used by the RunspacePool to send commands to the remote server, currently only supportsWSMan
apartment_state
: The int value ofpypsrp.complex_objects.ApartmentState
for the remote thread, default isUNKNOWN
thread_options
: The int value ofpypsrp.complex_objects.ThreadOptions
that specifies the type of thread to create, default isDEFAULT
host
: The local host info implementation, default is no hostconfiguration_name
: The configuration name to connect to, default isMicrosoft.PowerShell
and can be used to specify the Just Enough Administration (JEA) to connect tomin_runspaces
: The minimuum number of runspaces that a pool can hold, default is 1max_runspaces
: The maximum number of runspaces that a pool can hold. Each PowerShell pipeline is run in a single Runspace, default is 1session_key_timeout_ms
: The maximum time to wait for a session key transfer from the server
There are two process objects that can be used, pypsrp.shell.Process
for the
WinRS
shell and pypsrp.powershell.PowerShell
for the RunspacePool
shell.
These objects are ultimately used to execute commands, processes, or scripts on
the remote host.
Process
is used with the WinRS
shell to execute a cmd command or another
executable. The following options are used to configure the Process
object;
shell
: TheWinRS
shell the process is run overexecutable
: The executable or command to runarguments
: A list of arguments to the executable or command, default is no argumentsid
: The ID of the created command, if not specified then this is dynamically createdno_shell
: Whether to create a command in the cmd shell or bypass it, default isFalse
. IfTrue
then the executable must be the full path to the exe. This only works on older OS' before 2012 R2 (not including)
To execute the process, call .invoke()
, the stdout
, stderr
, and rc
properties contain the output of the command once complete.
PowerShell
is used by the PSRP protocol, it is designed to be a close
implementation of the
System.Management.Automation.PowerShell
class. The methods and properties are similar and can mostly do the same thing.
Here are the options that can be used to configure a PowerShell
process;
runspace_pool
: TheRunspacePool
object to run thePowerShell
process on
To execute the process, call .invoke()
, the output
, had_erros
, and
streams
contains the execution status and output information of the process.
Before invoke can be called, cmdlets or scripts must be added. These can be
done with the following methods;
add_script
: Add a raw PowerShell script to the pending commandsadd_cmdlet
: Add a cmdlet to the pending commandsadd_parameters
: Add a dictionary of key/value parameters to the last added commandadd_argument
: Add a value argument to the last added commandadd_statement
: Set the last command/script to be the end of that pipeline so the next command/script is like a newline
See the examples below for more details.
How to use the high level client API
from pypsrp.client import Client
# this takes in the same kwargs as the WSMan object
with Client("server", username="user", password="password") as client:
# execute a cmd command
stdout, stderr, rc = client.execute_cmd("dir")
stdout, stderr, rc = client.execute_cmd("powershell.exe gci $pwd")
sanitised_stderr = client.sanitise_clixml(stderr)
# execute a PowerShell script
output, streams, had_errors = client.execute_ps('''$path = "%s"
if (Test-Path -Path $path) {
Remove-Item -Path $path -Force -Recurse
}
New-Item -Path $path -ItemType Directory''' % path)
output, streams, had_errors = client.execute_ps("New-Item -Path C:\\temp\\folder -ItemType Directory")
# copy a file from the local host to the remote host
client.copy("~/file.txt", "C:\\temp\\file.txt")
# fetch a file from the remote host to the local host
client.fetch("C:\\temp\\file.txt", "~/file.txt")
How to use WinRS/Process to execute a command
from pypsrp.shell import Process, SignalCode, WinRS
from pypsrp.wsman import WSMan
# creates a http connection with no encryption and basic auth
wsman = WSMan("server", ssl=False, auth="basic", encryption="never",
username="vagrant", password="vagrant")
with wsman, WinRS(wsman) as shell:
process = Process(shell, "dir")
process.invoke()
process.signal(SignalCode.CTRL_C)
# execute a process with arguments in the background
process = Process(shell, "powershell", ["gci", "$pwd"])
process.begin_invoke() # start the invocation and return immediately
process.poll_invoke() # update the output stream
process.end_invoke() # finally wait until the process is finished
process.signal(SignalCode.CTRL_C)
How to use RunspacePool/PowerShell to execute a PowerShell script/command
from pypsrp.powershell import PowerShell, RunspacePool
from pypsrp.wsman import WSMan
# creates a https connection with explicit kerberos auth and implicit credentials
wsman = WSMan("server", auth="kerberos", cert_validation=False))
with wsman, RunspacePool(wsman) as pool:
# execute 'Get-Process | Select-Object Name'
ps = PowerShell(pool)
ps.add_cmdlet("Get-Process").add_cmdlet("Select-Object").add_argument("Name")
output = ps.invoke()
# execute 'Get-Process | Select-Object -Property Name'
ps.add_cmdlet("Get-Process").add_cmdlet("Select-Object")
ps.add_parameter("Property", "Name")
ps.begin_invoke() # execute process in the background
ps.poll_invoke() # update the output streams
ps.end_invoke() # wait until the process is finished
# execute 'Get-Process | Select-Object -Property Name; Get-Service audiosrv'
ps.add_cmdlet("Get-Process").add_cmdlet("Select-Object").add_parameter("Property", "Name")
ps.add_statement()
ps.add_cmdlet("Get-Service").add_argument("audiosrc")
ps.invoke()
# execute a PowerShell script with input being sent
script = '''begin {
$DebugPreference = "Continue"
Write-Debug -Message "begin"
} process {
Write-Output -InputObject $input
} end {
Write-Debug -Message "end"
}
'''
ps.add_script(script)
ps.invoke(["string", 1])
print(ps.output)
print(ps.streams.debug)
This library takes advantage of the Python logging configuration and messages
are logged to the pypsrp
named logger as well as pypsrp.*
where *
is each
Python script in the pypsrp
directory.
An easy way to turn on logging for the entire library is to create the
following JSON file and run your script with
PYPSRP_LOG_CFG=log.json python script.py
(this does not work with Python
2.6).
{
"version": 1,
"disable_existing_loggers": false,
"formatters": {
"simple": {
"format": "%(asctime)s - %(name)s - %(levelname)s - %(message)s"
}
},
"handlers": {
"console": {
"class": "logging.StreamHandler",
"level": "DEBUG",
"formatter": "simple",
"stream": "ext://sys.stdout"
}
},
"loggers": {
"pypsrp": {
"level": "DEBUG",
"handlers": ["console"],
"propagate": "no"
}
}
}
You can adjust the log level by changing the level value in logger
to INFO
.
Note: DEBUG
contains a lot of information and will output all the messages
sent to and from the client. This can have the side effect of leaking sensitive
information and should only be used for debugging purposes.
Any changes are more than welcome in pull request form, you can run the current test suite with tox like so;
# make sure tox is installed
pip install tox
# run the tox suite
tox
# or run the test manually for the current Python environment
py.test -v --pep8 --cov pypsrp --cov-report term-missing
A lot of the tests either simulate a remote Windows host but you can also run a lot of them against a real Windows host. To do this, set the following environment variables before running the tests;
PYPSRP_SERVER
: The hostname or IP of the remote hostPYPSRP_USERNAME
: The username to connect withPYPSRP_PASSWORD
: The password to connect withPYPSRR_PORT
: The port to connect with (default:5986
)PYPSRP_AUTH
: The authentication protocol to auth with (default:negotiate
)
There are further integration tests that require a specific host setup to run
correctly. You can use Vagrant
to set this host up. This is done by running
the following commands;
# download the Vagrant box and start it up based on the Vagrantfile
vagrant up
# once the above script is complete run the following
vagrant ssh # password is vagrant
powershell.exe
Register-PSSessionConfiguration -Path "C:\Users\vagrant\Documents\JEARoleSettings.pssc" -Name JEARole -Force
$sec_pass = ConvertTo-SecureString -String "vagrant" -AsPlainText -Force
$credential = New-Object -TypeName System.Management.Automation.PSCredential -ArgumentList "vagrant", $sec_pass
$thumbprint = (Get-ChildItem -Path Cert:\LocalMachine\TrustedPeople)[0].Thumbprint
New-Item -Path WSMan:\localhost\ClientCertificate `
-Subject "vagrant@localhost" `
-URI * `
-Issuer $thumbprint `
-Credential $credential `
-Force
# exit the remote PowerShell session
exit
# exist the SSH session
exit
Once complete, set the following environment variables to run the integration tests;
PYPSRP_RUN_INTEGRATION
: To any valuePYPSRP_SERVER
: Set to127.0.0.1
PYPSRP_USERNAME
: Set tovagrant
PYPSRP_PASSWORD
: Set tovagrant
PYPSRP_HTTP_PORT
: Set to55985
PYPSRP_HTTPS_PORT
: Set to55986
PYPSRP_CERT_DIR
: Set to the full path of the project directory
From here you can run the normal test suite and it will run all the integration tests.
- Look at implementing the following transport options
- Named pipes
- SSH
- Update CI to use named pipes for integration tests
- Add Ansible playbook for better integration tests
- Improved serialization between Python and .NET objects
- Live interactive console for PSRP