Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[WIP] LocalProcessProxy Kernels: user impersonate & working_dir #971

Open
wants to merge 4 commits into
base: main
Choose a base branch
from
Open

[WIP] LocalProcessProxy Kernels: user impersonate & working_dir #971

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
7c6d55c
Merge pull request #1 from jupyter/master
cceyda Feb 15, 2021
493da2b
Merge branch 'jupyter:master' into master
cceyda May 20, 2021
651f0b0
local user impersonation and working_dir
cceyda May 21, 2021
6bb1e03
set connection directory
cceyda May 21, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
17 changes: 14 additions & 3 deletions enterprise_gateway/services/kernels/remotemanager.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,7 +6,8 @@
import signal
import re
import uuid

from pwd import getpwnam
import grp
from tornado import web
from ipython_genutils.py3compat import unicode_type
from ipython_genutils.importstring import import_item
Expand All @@ -19,6 +20,7 @@
from enterprise_gateway.mixins import EnterpriseGatewayConfigMixin



def get_process_proxy_config(kernelspec):
"""
Return the process-proxy stanza from the kernelspec.
Expand Down Expand Up @@ -147,7 +149,7 @@ async def start_kernel(self, *args, **kwargs):
username = KernelSessionManager.get_kernel_username(**kwargs)
self.log.debug("RemoteMappingKernelManager.start_kernel: {kernel_name}, kernel_username: {username}".
format(kernel_name=kwargs['kernel_name'], username=username))

self.connection_dir=os.path.join(os.path.expanduser('~'+username),".local/share/jupyter/runtime/")
Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

self.connection_dir is normally determined by JUPYTER_RUNTIME_DIR but if this is not in user Paths in which to search for connection files it breaks so I did another hacky update. https://github.com/jupyter/jupyter_client/blob/98faaf91d00b4fbf17b2fbf0b08b1fd23adc27bf/jupyter_client/connect.py#L191

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Would it be better to call jupyter_core.paths.get_home_dir() to, presumably get /root, then replace that portion of jupyter_runtime_dir() with the target user's home directory? Otherwise, this isn't producing the correct path relative to the platform.

We should also ensure the user's home directory exists, since KERNEL_USERNAME is not necessarily reflective of a user. (I suppose setuid will enforce that constraint though.)

How will these kinds of files be cleaned up?

self.connection_dir is on the multi kernel manager which spans all kernel manager instances - so this isn't going to work. I'm not sure the connection information needs to be isolated on a per user level. It isn't today at least.

# Check max kernel limits
self._enforce_kernel_limits(username)

Expand Down Expand Up @@ -571,8 +573,17 @@ def write_connection_file(self):
self.hb_port = ports[3]
self.control_port = ports[4]

return super(RemoteKernelManager, self).write_connection_file()
cf = super(RemoteKernelManager, self).write_connection_file()
username=self.user_overrides.get('KERNEL_USERNAME',None)
Copy link
Author

@cceyda cceyda May 21, 2021 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is there a better way to get the username at this point? like:
username = KernelSessionManager.get_kernel_username(**kwargs)

It would be the best to write_connection_file to the correct user rather than chown-ing it later. But idk https://github.com/jupyter/jupyter_client/blob/98faaf91d00b4fbf17b2fbf0b08b1fd23adc27bf/jupyter_client/connect.py#L42

if username:
uid=getpwnam(username).pw_uid
guid=grp.getgrnam(username).gr_gid
os.chown(self.connection_file,uid,guid)
Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Had to change ownership of connection files. Because connection files are created by the gateway process (root) and when the kernel is launched as a user Permission denied happens. Alternative approaches are welcome~

self.log.debug(f"Local connection file chowned:{self.connection_file} to {username}")
else:
raise RuntimeError("Unrecognized user")

return cf
def _get_process_proxy(self):
"""
Reads the associated kernelspec and to see if has a process proxy stanza.
Expand Down
15 changes: 13 additions & 2 deletions enterprise_gateway/services/processproxies/processproxy.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,7 +36,7 @@
from zmq.ssh import tunnel

from ..sessions.kernelsessionmanager import KernelSessionManager

from jupyterhub.spawner import set_user_setuid
# Default logging level of paramiko produces too much noise - raise to warning only.
logging.getLogger('paramiko').setLevel(os.getenv('EG_SSH_LOG_LEVEL', logging.WARNING))

Expand Down Expand Up @@ -911,7 +911,18 @@ def __init__(self, kernel_manager, proxy_config):

async def launch_process(self, kernel_cmd, **kwargs):
await super(LocalProcessProxy, self).launch_process(kernel_cmd, **kwargs)

user=kwargs["env"].get("KERNEL_USERNAME",None)
self.log.info(f"KARGS {kwargs}")

if user:
Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can check here if EG_IMPERSONATION_ENABLED & user. (The current EG behavior ignores the KERNEL_USERNAME)

btw,I will also remove unnecessary logging, it was to help me understand how EG works.

kwargs["preexec_fn"]=set_user_setuid(user,chdir=False)
else:
raise RuntimeError("Trying to run as root on host, unrecognized user")
# launch the local run.sh
working_dir=kwargs["env"].get("KERNEL_WORKING_DIR",None)
if working_dir:
kwargs["cwd"]=working_dir

# launch the local run.sh
self.local_proc = self.launch_kernel(kernel_cmd, **kwargs)
self.pid = self.local_proc.pid
Expand Down