##// END OF EJS Templates
deps: bumped pycryptodome==3.21.0 for security issue
deps: bumped pycryptodome==3.21.0 for security issue

File last commit:

r5608:6d33e504 default
r5640:acc4336c default
Show More
simplevcs.py
659 lines | 26.4 KiB | text/x-python | PythonLexer
# Copyright (C) 2014-2024 RhodeCode GmbH
#
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU Affero General Public License, version 3
# (only), as published by the Free Software Foundation.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU Affero General Public License
# along with this program. If not, see <http://www.gnu.org/licenses/>.
#
# This program is dual-licensed. If you wish to learn more about the
# RhodeCode Enterprise Edition, including its added features, Support services,
# and proprietary license terms, please see https://rhodecode.com/licenses/
"""
SimpleVCS middleware for handling protocol request (push/clone etc.)
It's implemented with basic auth function
"""
import os
import re
import logging
import importlib
from functools import wraps
import time
from paste.httpheaders import REMOTE_USER, AUTH_TYPE
from pyramid.httpexceptions import HTTPNotFound, HTTPForbidden, HTTPNotAcceptable, HTTPInternalServerError
from zope.cachedescriptors.property import Lazy as LazyProperty
import rhodecode
from rhodecode.authentication.base import authenticate, VCS_TYPE, loadplugin
from rhodecode.lib import rc_cache
from rhodecode.lib.svn_txn_utils import store_txn_id_data
from rhodecode.lib.auth import AuthUser, HasPermissionAnyMiddleware
from rhodecode.lib.base import BasicAuth, get_ip_addr, get_user_agent, vcs_operation_context
from rhodecode.lib.exceptions import UserCreationError, NotAllowedToCreateUserError
from rhodecode.lib.hook_daemon.utils import prepare_callback_daemon
from rhodecode.lib.middleware import appenlight
from rhodecode.lib.middleware.utils import scm_app_http
from rhodecode.lib.str_utils import safe_bytes, safe_int
from rhodecode.lib.utils import is_valid_repo, SLUG_RE
from rhodecode.lib.utils2 import safe_str, fix_PATH, str2bool
from rhodecode.lib.vcs.conf import settings as vcs_settings
from rhodecode.lib.vcs.backends import base
from rhodecode.model import meta
from rhodecode.model.db import User, Repository, PullRequest
from rhodecode.model.scm import ScmModel
from rhodecode.model.pull_request import PullRequestModel
from rhodecode.model.settings import SettingsModel, VcsSettingsModel
log = logging.getLogger(__name__)
def initialize_generator(factory):
"""
Initializes the returned generator by draining its first element.
This can be used to give a generator an initializer, which is the code
up to the first yield statement. This decorator enforces that the first
produced element has the value ``"__init__"`` to make its special
purpose very explicit in the using code.
"""
@wraps(factory)
def wrapper(*args, **kwargs):
gen = factory(*args, **kwargs)
try:
init = next(gen)
except StopIteration:
raise ValueError("Generator must yield at least one element.")
if init != "__init__":
raise ValueError('First yielded element must be "__init__".')
return gen
return wrapper
class SimpleVCS(object):
"""Common functionality for SCM HTTP handlers."""
SCM = "unknown"
acl_repo_name = None
url_repo_name = None
vcs_repo_name = None
rc_extras = {}
# We have to handle requests to shadow repositories different than requests
# to normal repositories. Therefore we have to distinguish them. To do this
# we use this regex which will match only on URLs pointing to shadow
# repositories.
shadow_repo_re = re.compile(
"(?P<groups>(?:{slug_pat}/)*)" # repo groups
"(?P<target>{slug_pat})/" # target repo
"pull-request/(?P<pr_id>\\d+)/" # pull request
"repository$".format(slug_pat=SLUG_RE.pattern) # shadow repo
)
def __init__(self, config, registry):
self.registry = registry
self.config = config
# re-populated by specialized middleware
self.repo_vcs_config = base.Config()
rc_settings = SettingsModel().get_all_settings(cache=True, from_request=False)
realm = rc_settings.get("rhodecode_realm") or "RhodeCode AUTH"
# authenticate this VCS request using authfunc
auth_ret_code_detection = str2bool(self.config.get("auth_ret_code_detection", False))
self.authenticate = BasicAuth(
"", authenticate, registry, config.get("auth_ret_code"), auth_ret_code_detection, rc_realm=realm
)
self.ip_addr = "0.0.0.0"
@LazyProperty
def global_vcs_config(self):
try:
return VcsSettingsModel().get_ui_settings_as_config_obj()
except Exception:
return base.Config()
@property
def base_path(self):
settings_path = self.config.get("repo_store.path")
if not settings_path:
raise ValueError("FATAL: repo_store.path is empty")
return settings_path
def set_repo_names(self, environ):
"""
This will populate the attributes acl_repo_name, url_repo_name,
vcs_repo_name and is_shadow_repo. In case of requests to normal (non
shadow) repositories all names are equal. In case of requests to a
shadow repository the acl-name points to the target repo of the pull
request and the vcs-name points to the shadow repo file system path.
The url-name is always the URL used by the vcs client program.
Example in case of a shadow repo:
acl_repo_name = RepoGroup/MyRepo
url_repo_name = RepoGroup/MyRepo/pull-request/3/repository
vcs_repo_name = /repo/base/path/RepoGroup/.__shadow_MyRepo_pr-3'
"""
# First we set the repo name from URL for all attributes. This is the
# default if handling normal (non shadow) repo requests.
self.url_repo_name = self._get_repository_name(environ)
self.acl_repo_name = self.vcs_repo_name = self.url_repo_name
self.is_shadow_repo = False
# Check if this is a request to a shadow repository.
match = self.shadow_repo_re.match(self.url_repo_name)
if match:
match_dict = match.groupdict()
# Build acl repo name from regex match.
acl_repo_name = safe_str(
"{groups}{target}".format(groups=match_dict["groups"] or "", target=match_dict["target"])
)
# Retrieve pull request instance by ID from regex match.
pull_request = PullRequest.get(match_dict["pr_id"])
# Only proceed if we got a pull request and if acl repo name from
# URL equals the target repo name of the pull request.
if pull_request and (acl_repo_name == pull_request.target_repo.repo_name):
# Get file system path to shadow repository.
workspace_id = PullRequestModel()._workspace_id(pull_request)
vcs_repo_name = pull_request.target_repo.get_shadow_repository_path(workspace_id)
# Store names for later usage.
self.vcs_repo_name = vcs_repo_name
self.acl_repo_name = acl_repo_name
self.is_shadow_repo = True
log.debug(
"Setting all VCS repository names: %s",
{
"acl_repo_name": self.acl_repo_name,
"url_repo_name": self.url_repo_name,
"vcs_repo_name": self.vcs_repo_name,
},
)
@property
def scm_app(self):
custom_implementation = self.config["vcs.scm_app_implementation"]
if custom_implementation == "http":
log.debug("Using HTTP implementation of scm app.")
scm_app_impl = scm_app_http
else:
log.debug('Using custom implementation of scm_app: "{}"'.format(custom_implementation))
scm_app_impl = importlib.import_module(custom_implementation)
return scm_app_impl
def _get_by_id(self, repo_name):
"""
Gets a special pattern _<ID> from clone url and tries to replace it
with a repository_name for support of _<ID> non changeable urls
"""
data = repo_name.split("/")
if len(data) >= 2:
from rhodecode.model.repo import RepoModel
by_id_match = RepoModel().get_repo_by_id(repo_name)
if by_id_match:
data[1] = by_id_match.repo_name
# Because PEP-3333-WSGI uses bytes-tunneled-in-latin-1 as PATH_INFO
# and we use this data
maybe_new_path = "/".join(data)
return safe_bytes(maybe_new_path).decode("latin1")
def _invalidate_cache(self, repo_name):
"""
Set's cache for this repository for invalidation on next access
:param repo_name: full repo name, also a cache key
"""
ScmModel().mark_for_invalidation(repo_name)
def is_valid_and_existing_repo(self, repo_name, base_path, scm_type):
db_repo = Repository.get_by_repo_name(repo_name)
if not db_repo:
log.debug("Repository `%s` not found inside the database.", repo_name)
return False
if db_repo.repo_type != scm_type:
log.warning(
"Repository `%s` have incorrect scm_type, expected %s got %s", repo_name, db_repo.repo_type, scm_type
)
return False
config = db_repo._config
config.set("extensions", "largefiles", "")
return is_valid_repo(repo_name, base_path, explicit_scm=scm_type, expect_scm=scm_type, config=config)
def valid_and_active_user(self, user):
"""
Checks if that user is not empty, and if it's actually object it checks
if he's active.
:param user: user object or None
:return: boolean
"""
if user is None:
return False
elif user.active:
return True
return False
@property
def is_shadow_repo_dir(self):
return os.path.isdir(self.vcs_repo_name)
def _check_permission(
self, action, user, auth_user, repo_name, ip_addr=None, plugin_id="", plugin_cache_active=False, cache_ttl=0
):
"""
Checks permissions using action (push/pull) user and repository
name. If plugin_cache and ttl is set it will use the plugin which
authenticated the user to store the cached permissions result for N
amount of seconds as in cache_ttl
:param action: push or pull action
:param user: user instance
:param repo_name: repository name
"""
log.debug("AUTH_CACHE_TTL for permissions `%s` active: %s (TTL: %s)", plugin_id, plugin_cache_active, cache_ttl)
user_id = user.user_id
cache_namespace_uid = f"cache_user_auth.{rc_cache.PERMISSIONS_CACHE_VER}.{user_id}"
region = rc_cache.get_or_create_region("cache_perms", cache_namespace_uid)
@region.conditional_cache_on_arguments(
namespace=cache_namespace_uid, expiration_time=cache_ttl, condition=plugin_cache_active
)
def compute_perm_vcs(cache_name, plugin_id, action, user_id, repo_name, ip_addr):
log.debug("auth: calculating permission access now for vcs operation: %s", action)
# check IP
inherit = user.inherit_default_permissions
ip_allowed = AuthUser.check_ip_allowed(user_id, ip_addr, inherit_from_default=inherit)
if ip_allowed:
log.info("Access for IP:%s allowed", ip_addr)
else:
return False
if action == "push":
perms = ("repository.write", "repository.admin")
if not HasPermissionAnyMiddleware(*perms)(auth_user, repo_name):
return False
else:
# any other action need at least read permission
perms = ("repository.read", "repository.write", "repository.admin")
if not HasPermissionAnyMiddleware(*perms)(auth_user, repo_name):
return False
return True
start = time.time()
log.debug("Running plugin `%s` permissions check", plugin_id)
# for environ based auth, password can be empty, but then the validation is
# on the server that fills in the env data needed for authentication
perm_result = compute_perm_vcs("vcs_permissions", plugin_id, action, user.user_id, repo_name, ip_addr)
auth_time = time.time() - start
log.debug(
"Permissions for plugin `%s` completed in %.4fs, " "expiration time of fetched cache %.1fs.",
plugin_id,
auth_time,
cache_ttl,
)
return perm_result
def _get_http_scheme(self, environ):
try:
return environ["wsgi.url_scheme"]
except Exception:
log.exception("Failed to read http scheme")
return "http"
def _get_default_cache_ttl(self):
# take AUTH_CACHE_TTL from the `rhodecode` auth plugin
plugin = loadplugin("egg:rhodecode-enterprise-ce#rhodecode")
plugin_settings = plugin.get_settings()
plugin_cache_active, cache_ttl = plugin.get_ttl_cache(plugin_settings) or (False, 0)
return plugin_cache_active, cache_ttl
def __call__(self, environ, start_response):
try:
return self._handle_request(environ, start_response)
except Exception:
log.exception("Exception while handling request")
appenlight.track_exception(environ)
return HTTPInternalServerError()(environ, start_response)
finally:
meta.Session.remove()
def _handle_request(self, environ, start_response):
if not self.url_repo_name:
log.warning("Repository name is empty: %s", self.url_repo_name)
# failed to get repo name, we fail now
return HTTPNotFound()(environ, start_response)
log.debug("Extracted repo name is %s", self.url_repo_name)
ip_addr = get_ip_addr(environ)
user_agent = get_user_agent(environ)
username = None
# skip passing error to error controller
environ["pylons.status_code_redirect"] = True
# ======================================================================
# GET ACTION PULL or PUSH
# ======================================================================
action = self._get_action(environ)
# ======================================================================
# Check if this is a request to a shadow repository of a pull request.
# In this case only pull action is allowed.
# ======================================================================
if self.is_shadow_repo and action != "pull":
reason = "Only pull action is allowed for shadow repositories."
log.debug("User not allowed to proceed, %s", reason)
return HTTPNotAcceptable(reason)(environ, start_response)
# Check if the shadow repo actually exists, in case someone refers
# to it, and it has been deleted because of successful merge.
if self.is_shadow_repo and not self.is_shadow_repo_dir:
log.debug("Shadow repo detected, and shadow repo dir `%s` is missing", self.is_shadow_repo_dir)
return HTTPNotFound()(environ, start_response)
# ======================================================================
# CHECK ANONYMOUS PERMISSION
# ======================================================================
detect_force_push = False
check_branch_perms = False
if action in ["pull", "push"]:
user_obj = anonymous_user = User.get_default_user()
auth_user = user_obj.AuthUser()
username = anonymous_user.username
if anonymous_user.active:
plugin_cache_active, cache_ttl = self._get_default_cache_ttl()
# ONLY check permissions if the user is activated
anonymous_perm = self._check_permission(
action,
anonymous_user,
auth_user,
self.acl_repo_name,
ip_addr,
plugin_id="anonymous_access",
plugin_cache_active=plugin_cache_active,
cache_ttl=cache_ttl,
)
else:
anonymous_perm = False
if not anonymous_user.active or not anonymous_perm:
if not anonymous_user.active:
log.debug("Anonymous access is disabled, running " "authentication")
if not anonymous_perm:
log.debug(
"Not enough credentials to access repo: `%s` " "repository as anonymous user",
self.acl_repo_name,
)
username = None
# ==============================================================
# DEFAULT PERM FAILED OR ANONYMOUS ACCESS IS DISABLED SO WE
# NEED TO AUTHENTICATE AND ASK FOR AUTH USER PERMISSIONS
# ==============================================================
# try to auth based on environ, container auth methods
log.debug("Running PRE-AUTH for container|headers based authentication")
# headers auth, by just reading special headers and bypass the auth with user/passwd
pre_auth = authenticate(
"", "", environ, VCS_TYPE, registry=self.registry, acl_repo_name=self.acl_repo_name
)
if pre_auth and pre_auth.get("username"):
username = pre_auth["username"]
log.debug("PRE-AUTH got `%s` as username", username)
if pre_auth:
log.debug("PRE-AUTH successful from %s", pre_auth.get("auth_data", {}).get("_plugin"))
# If not authenticated by the container, running basic auth
# before inject the calling repo_name for special scope checks
self.authenticate.acl_repo_name = self.acl_repo_name
plugin_cache_active, cache_ttl = False, 0
plugin = None
# regular auth chain
if not username:
self.authenticate.realm = self.authenticate.get_rc_realm()
try:
auth_result = self.authenticate(environ)
except (UserCreationError, NotAllowedToCreateUserError) as e:
log.error(e)
reason = safe_str(e)
return HTTPNotAcceptable(reason)(environ, start_response)
if isinstance(auth_result, dict):
AUTH_TYPE.update(environ, "basic")
REMOTE_USER.update(environ, auth_result["username"])
username = auth_result["username"]
plugin = auth_result.get("auth_data", {}).get("_plugin")
log.info("MAIN-AUTH successful for user `%s` from %s plugin", username, plugin)
plugin_cache_active, cache_ttl = auth_result.get("auth_data", {}).get("_ttl_cache") or (
False,
0,
)
else:
return auth_result.wsgi_application(environ, start_response)
# ==============================================================
# CHECK PERMISSIONS FOR THIS REQUEST USING GIVEN USERNAME
# ==============================================================
user = User.get_by_username(username)
if not self.valid_and_active_user(user):
return HTTPForbidden()(environ, start_response)
username = user.username
user_id = user.user_id
# check user attributes for password change flag
user_obj = user
auth_user = user_obj.AuthUser()
if (
user_obj
and user_obj.username != User.DEFAULT_USER
and user_obj.user_data.get("force_password_change")
):
reason = "password change required"
log.debug("User not allowed to authenticate, %s", reason)
return HTTPNotAcceptable(reason)(environ, start_response)
# check permissions for this repository
perm = self._check_permission(
action, user, auth_user, self.acl_repo_name, ip_addr, plugin, plugin_cache_active, cache_ttl
)
if not perm:
return HTTPForbidden()(environ, start_response)
environ["rc_auth_user_id"] = str(user_id)
if action == "push":
perms = auth_user.get_branch_permissions(self.acl_repo_name)
if perms:
check_branch_perms = True
detect_force_push = True
# extras are injected into UI object and later available
# in hooks executed by RhodeCode
check_locking = _should_check_locking(environ.get("QUERY_STRING"))
extras = vcs_operation_context(
environ,
repo_name=self.acl_repo_name,
username=username,
action=action,
scm=self.SCM,
check_locking=check_locking,
is_shadow_repo=self.is_shadow_repo,
check_branch_perms=check_branch_perms,
detect_force_push=detect_force_push,
)
# ======================================================================
# REQUEST HANDLING
# ======================================================================
repo_path = os.path.join(safe_str(self.base_path), safe_str(self.vcs_repo_name))
log.debug("Repository path is %s", repo_path)
fix_PATH()
log.info(
'%s action on %s repo "%s" by "%s" from %s %s',
action,
self.SCM,
safe_str(self.url_repo_name),
safe_str(username),
ip_addr,
user_agent,
)
return self._generate_vcs_response(environ, start_response, repo_path, extras, action)
def _get_txn_id(self, environ):
for k in ["RAW_URI", "HTTP_DESTINATION"]:
url = environ.get(k)
if not url:
continue
# regex to search for svn-txn-id
pattern = r"/!svn/txr/([^/]+)/"
# Search for the pattern in the URL
match = re.search(pattern, url)
# Check if a match is found and extract the captured group
if match:
txn_id = match.group(1)
return txn_id
@initialize_generator
def _generate_vcs_response(self, environ, start_response, repo_path, extras, action):
"""
Returns a generator for the response content.
This method is implemented as a generator, so that it can trigger
the cache validation after all content sent back to the client. It
also handles the locking exceptions which will be triggered when
the first chunk is produced by the underlying WSGI application.
"""
svn_txn_id = ""
if action == "push":
svn_txn_id = self._get_txn_id(environ)
callback_daemon, extras = self._prepare_callback_daemon(extras, environ, action, txn_id=svn_txn_id)
if svn_txn_id:
txn_id_data = extras.copy()
txn_id_data.update({"req_method": environ["REQUEST_METHOD"]})
full_repo_path = repo_path
store_txn_id_data(full_repo_path, svn_txn_id, txn_id_data)
log.debug("HOOKS extras is %s", extras)
http_scheme = self._get_http_scheme(environ)
config = self._create_config(extras, self.acl_repo_name, scheme=http_scheme)
app = self._create_wsgi_app(repo_path, self.url_repo_name, config)
with callback_daemon:
app.rc_extras = extras
try:
response = app(environ, start_response)
finally:
# This statement works together with the decorator
# "initialize_generator" above. The decorator ensures that
# we hit the first yield statement before the generator is
# returned back to the WSGI server. This is needed to
# ensure that the call to "app" above triggers the
# needed callback to "start_response" before the
# generator is actually used.
yield "__init__"
# iter content
for chunk in response:
yield chunk
try:
# invalidate cache on push
if action == "push":
self._invalidate_cache(self.url_repo_name)
finally:
meta.Session.remove()
def _get_repository_name(self, environ):
"""Get repository name out of the environmnent
:param environ: WSGI environment
"""
raise NotImplementedError()
def _get_action(self, environ):
"""Map request commands into a pull or push command.
:param environ: WSGI environment
"""
raise NotImplementedError()
def _create_wsgi_app(self, repo_path, repo_name, config):
"""Return the WSGI app that will finally handle the request."""
raise NotImplementedError()
def _create_config(self, extras, repo_name, scheme="http"):
"""Create a safe config representation."""
raise NotImplementedError()
def _should_use_callback_daemon(self, extras, environ, action):
if extras.get("is_shadow_repo"):
# we don't want to execute hooks, and callback daemon for shadow repos
return False
return True
def _prepare_callback_daemon(self, extras, environ, action, txn_id=None):
protocol = vcs_settings.HOOKS_PROTOCOL
if not self._should_use_callback_daemon(extras, environ, action):
# disable callback daemon for actions that don't require it
protocol = "local"
return prepare_callback_daemon(extras, protocol=protocol, txn_id=txn_id)
def _should_check_locking(query_string):
# this is kind of hacky, but due to how mercurial handles client-server
# server see all operation on commit; bookmarks, phases and
# obsolescence marker in different transaction, we don't want to check
# locking on those
return query_string not in ["cmd=listkeys"]