Changeset - 21084a951cd9
[Not reviewed]
stable
0 2 0
Mads Kiilerich - 7 years ago 2018-10-11 02:06:50
mads@kiilerich.com
Grafted from: 61668afbc59b
hg: make __get_action command parsing simpler and safer
2 files changed with 6 insertions and 9 deletions:
0 comments (0 inline, 0 general)
kallithea/lib/base.py
Show inline comments
 
@@ -29,385 +29,385 @@ Original author and date, and relevant c
 
"""
 

	
 
import datetime
 
import logging
 
import time
 
import traceback
 

	
 
import webob.exc
 
import paste.httpexceptions
 
import paste.auth.basic
 
import paste.httpheaders
 

	
 
from pylons import config, tmpl_context as c, request, session, url
 
from pylons.controllers import WSGIController
 
from pylons.controllers.util import redirect
 
from pylons.templating import render_mako as render  # don't remove this import
 
from pylons.i18n.translation import _
 

	
 
from kallithea import __version__, BACKENDS
 

	
 
from kallithea.lib.utils2 import str2bool, safe_unicode, AttributeDict,\
 
    safe_str, safe_int
 
from kallithea.lib import auth_modules
 
from kallithea.lib.auth import AuthUser, HasPermissionAnyMiddleware
 
from kallithea.lib.utils import get_repo_slug
 
from kallithea.lib.exceptions import UserCreationError
 
from kallithea.lib.vcs.exceptions import RepositoryError, EmptyRepositoryError, ChangesetDoesNotExistError
 
from kallithea.model import meta
 

	
 
from kallithea.model.db import Repository, Ui, User, Setting
 
from kallithea.model.notification import NotificationModel
 
from kallithea.model.scm import ScmModel
 
from kallithea.model.pull_request import PullRequestModel
 

	
 
log = logging.getLogger(__name__)
 

	
 

	
 
def _filter_proxy(ip):
 
    """
 
    HEADERS can have multiple ips inside the left-most being the original
 
    client, and each successive proxy that passed the request adding the IP
 
    address where it received the request from.
 

	
 
    :param ip:
 
    """
 
    if ',' in ip:
 
        _ips = ip.split(',')
 
        _first_ip = _ips[0].strip()
 
        log.debug('Got multiple IPs %s, using %s', ','.join(_ips), _first_ip)
 
        return _first_ip
 
    return ip
 

	
 

	
 
def _get_ip_addr(environ):
 
    proxy_key = 'HTTP_X_REAL_IP'
 
    proxy_key2 = 'HTTP_X_FORWARDED_FOR'
 
    def_key = 'REMOTE_ADDR'
 

	
 
    ip = environ.get(proxy_key)
 
    if ip:
 
        return _filter_proxy(ip)
 

	
 
    ip = environ.get(proxy_key2)
 
    if ip:
 
        return _filter_proxy(ip)
 

	
 
    ip = environ.get(def_key, '0.0.0.0')
 
    return _filter_proxy(ip)
 

	
 

	
 
def _get_access_path(environ):
 
    path = environ.get('PATH_INFO')
 
    org_req = environ.get('pylons.original_request')
 
    if org_req:
 
        path = org_req.environ.get('PATH_INFO')
 
    return path
 

	
 

	
 
def log_in_user(user, remember, is_external_auth):
 
    """
 
    Log a `User` in and update session and cookies. If `remember` is True,
 
    the session cookie is set to expire in a year; otherwise, it expires at
 
    the end of the browser session.
 

	
 
    Returns populated `AuthUser` object.
 
    """
 
    user.update_lastlogin()
 
    meta.Session().commit()
 

	
 
    auth_user = AuthUser(dbuser=user,
 
                         is_external_auth=is_external_auth)
 
    auth_user.set_authenticated()
 

	
 
    # Start new session to prevent session fixation attacks.
 
    session.invalidate()
 
    session['authuser'] = cookie = auth_user.to_cookie()
 

	
 
    # If they want to be remembered, update the cookie.
 
    # NOTE: Assumes that beaker defaults to browser session cookie.
 
    if remember:
 
        t = datetime.datetime.now() + datetime.timedelta(days=365)
 
        session._set_cookie_expires(t)
 

	
 
    session.save()
 

	
 
    log.info('user %s is now authenticated and stored in '
 
             'session, session attrs %s', user.username, cookie)
 

	
 
    # dumps session attrs back to cookie
 
    session._update_cookie_out()
 

	
 
    return auth_user
 

	
 

	
 
class BasicAuth(paste.auth.basic.AuthBasicAuthenticator):
 

	
 
    def __init__(self, realm, authfunc, auth_http_code=None):
 
        self.realm = realm
 
        self.authfunc = authfunc
 
        self._rc_auth_http_code = auth_http_code
 

	
 
    def build_authentication(self):
 
        head = paste.httpheaders.WWW_AUTHENTICATE.tuples('Basic realm="%s"' % self.realm)
 
        if self._rc_auth_http_code and self._rc_auth_http_code == '403':
 
            # return 403 if alternative http return code is specified in
 
            # Kallithea config
 
            return paste.httpexceptions.HTTPForbidden(headers=head)
 
        return paste.httpexceptions.HTTPUnauthorized(headers=head)
 

	
 
    def authenticate(self, environ):
 
        authorization = paste.httpheaders.AUTHORIZATION(environ)
 
        if not authorization:
 
            return self.build_authentication()
 
        (authmeth, auth) = authorization.split(' ', 1)
 
        if 'basic' != authmeth.lower():
 
            return self.build_authentication()
 
        auth = auth.strip().decode('base64')
 
        _parts = auth.split(':', 1)
 
        if len(_parts) == 2:
 
            username, password = _parts
 
            if self.authfunc(username, password, environ) is not None:
 
                return username
 
        return self.build_authentication()
 

	
 
    __call__ = authenticate
 

	
 

	
 
class BaseVCSController(object):
 

	
 
    def __init__(self, application, config):
 
        self.application = application
 
        self.config = config
 
        # base path of repo locations
 
        self.basepath = self.config['base_path']
 
        # authenticate this VCS request using the authentication modules
 
        self.authenticate = BasicAuth('', auth_modules.authenticate,
 
                                      config.get('auth_ret_code'))
 
        self.ip_addr = '0.0.0.0'
 

	
 
    def _handle_request(self, environ, start_response):
 
        raise NotImplementedError()
 

	
 
    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> permanent URLs
 

	
 
        :param repo_name:
 
        """
 

	
 
        data = repo_name.split('/')
 
        if len(data) >= 2:
 
            from kallithea.lib.utils import get_repo_by_id
 
            by_id_match = get_repo_by_id(repo_name)
 
            if by_id_match:
 
                data[1] = safe_str(by_id_match)
 

	
 
        return '/'.join(data)
 

	
 
    def _invalidate_cache(self, repo_name):
 
        """
 
        Sets 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 _check_permission(self, action, user, repo_name, ip_addr=None):
 
        """
 
        Checks permissions using action (push/pull) user and repository
 
        name
 

	
 
        :param action: push or pull action
 
        :param action: 'push' or 'pull' action
 
        :param user: `User` instance
 
        :param repo_name: repository name
 
        """
 
        # check IP
 
        ip_allowed = AuthUser.check_ip_allowed(user, ip_addr)
 
        if ip_allowed:
 
            log.info('Access for IP:%s allowed', ip_addr)
 
        else:
 
            return False
 

	
 
        if action == 'push':
 
            if not HasPermissionAnyMiddleware('repository.write',
 
                                              'repository.admin')(user,
 
                                                                  repo_name):
 
                return False
 

	
 
        else:
 
            #any other action need at least read permission
 
            if not HasPermissionAnyMiddleware('repository.read',
 
                                              'repository.write',
 
                                              'repository.admin')(user,
 
                                                                  repo_name):
 
                return False
 

	
 
        return True
 

	
 
    def _get_ip_addr(self, environ):
 
        return _get_ip_addr(environ)
 

	
 
    def _check_ssl(self, environ):
 
        """
 
        Checks the SSL check flag and returns False if SSL is not present
 
        and required True otherwise
 
        """
 
        #check if we have SSL required  ! if not it's a bad request !
 
        if str2bool(Ui.get_by_key('push_ssl').ui_value):
 
            org_proto = environ.get('wsgi._org_proto', environ['wsgi.url_scheme'])
 
            if org_proto != 'https':
 
                log.debug('proto is %s and SSL is required BAD REQUEST !',
 
                          org_proto)
 
                return False
 
        return True
 

	
 
    def _check_locking_state(self, environ, action, repo, user_id):
 
        """
 
        Checks locking on this repository, if locking is enabled and lock is
 
        present returns a tuple of make_lock, locked, locked_by.
 
        make_lock can have 3 states None (do nothing) True, make lock
 
        False release lock, This value is later propagated to hooks, which
 
        do the locking. Think about this as signals passed to hooks what to do.
 

	
 
        """
 
        locked = False  # defines that locked error should be thrown to user
 
        make_lock = None
 
        repo = Repository.get_by_repo_name(repo)
 
        user = User.get(user_id)
 

	
 
        # this is kind of hacky, but due to how mercurial handles client-server
 
        # server see all operation on changeset; bookmarks, phases and
 
        # obsolescence marker in different transaction, we don't want to check
 
        # locking on those
 
        obsolete_call = environ['QUERY_STRING'] in ['cmd=listkeys',]
 
        locked_by = repo.locked
 
        if repo and repo.enable_locking and not obsolete_call:
 
            if action == 'push':
 
                #check if it's already locked !, if it is compare users
 
                user_id, _date = repo.locked
 
                if user.user_id == user_id:
 
                    log.debug('Got push from user %s, now unlocking', user)
 
                    # unlock if we have push from user who locked
 
                    make_lock = False
 
                else:
 
                    # we're not the same user who locked, ban with 423 !
 
                    locked = True
 
            if action == 'pull':
 
                if repo.locked[0] and repo.locked[1]:
 
                    locked = True
 
                else:
 
                    log.debug('Setting lock on repo %s by %s', repo, user)
 
                    make_lock = True
 

	
 
        else:
 
            log.debug('Repository %s do not have locking enabled', repo)
 
        log.debug('FINAL locking values make_lock:%s,locked:%s,locked_by:%s',
 
                  make_lock, locked, locked_by)
 
        return make_lock, locked, locked_by
 

	
 
    def __call__(self, environ, start_response):
 
        start = time.time()
 
        try:
 
            return self._handle_request(environ, start_response)
 
        finally:
 
            log = logging.getLogger('kallithea.' + self.__class__.__name__)
 
            log.debug('Request time: %.3fs', time.time() - start)
 
            meta.Session.remove()
 

	
 

	
 
class BaseController(WSGIController):
 

	
 
    def __before__(self):
 
        """
 
        __before__ is called before controller methods and after __call__
 
        """
 
        c.kallithea_version = __version__
 
        rc_config = Setting.get_app_settings()
 

	
 
        # Visual options
 
        c.visual = AttributeDict({})
 

	
 
        ## DB stored
 
        c.visual.show_public_icon = str2bool(rc_config.get('show_public_icon'))
 
        c.visual.show_private_icon = str2bool(rc_config.get('show_private_icon'))
 
        c.visual.stylify_metatags = str2bool(rc_config.get('stylify_metatags'))
 
        c.visual.dashboard_items = safe_int(rc_config.get('dashboard_items', 100))
 
        c.visual.admin_grid_items = safe_int(rc_config.get('admin_grid_items', 100))
 
        c.visual.repository_fields = str2bool(rc_config.get('repository_fields'))
 
        c.visual.show_version = str2bool(rc_config.get('show_version'))
 
        c.visual.use_gravatar = str2bool(rc_config.get('use_gravatar'))
 
        c.visual.gravatar_url = rc_config.get('gravatar_url')
 

	
 
        c.ga_code = rc_config.get('ga_code')
 
        # TODO: replace undocumented backwards compatibility hack with db upgrade and rename ga_code
 
        if c.ga_code and '<' not in c.ga_code:
 
            c.ga_code = '''<script type="text/javascript">
 
                var _gaq = _gaq || [];
 
                _gaq.push(['_setAccount', '%s']);
 
                _gaq.push(['_trackPageview']);
 

	
 
                (function() {
 
                    var ga = document.createElement('script'); ga.type = 'text/javascript'; ga.async = true;
 
                    ga.src = ('https:' == document.location.protocol ? 'https://ssl' : 'http://www') + '.google-analytics.com/ga.js';
 
                    var s = document.getElementsByTagName('script')[0]; s.parentNode.insertBefore(ga, s);
 
                    })();
 
            </script>''' % c.ga_code
 
        c.site_name = rc_config.get('title')
 
        c.clone_uri_tmpl = rc_config.get('clone_uri_tmpl')
 

	
 
        ## INI stored
 
        c.visual.allow_repo_location_change = str2bool(config.get('allow_repo_location_change', True))
 
        c.visual.allow_custom_hooks_settings = str2bool(config.get('allow_custom_hooks_settings', True))
 

	
 
        c.instance_id = config.get('instance_id')
 
        c.issues_url = config.get('bugtracker', url('issues_url'))
 
        # END CONFIG VARS
 

	
 
        c.repo_name = get_repo_slug(request)  # can be empty
 
        c.backends = BACKENDS.keys()
 
        c.unread_notifications = NotificationModel()\
 
                        .get_unread_cnt_for_user(c.authuser.user_id)
 

	
 
        self.cut_off_limit = safe_int(config.get('cut_off_limit'))
 

	
 
        c.my_pr_count = PullRequestModel().get_pullrequest_cnt_for_user(c.authuser.user_id)
 

	
 
        self.sa = meta.Session
 
        self.scm_model = ScmModel(self.sa)
 

	
 
    @staticmethod
 
    def _determine_auth_user(api_key, session_authuser):
 
        """
 
        Create an `AuthUser` object given the API key (if any) and the
 
        value of the authuser session cookie.
 
        """
 

	
 
        # Authenticate by API key
 
        if api_key:
 
            # when using API_KEY we are sure user exists.
 
            return AuthUser(dbuser=User.get_by_api_key(api_key),
 
                            is_external_auth=True)
 

	
 
        # Authenticate by session cookie
 
        # In ancient login sessions, 'authuser' may not be a dict.
 
        # In that case, the user will have to log in again.
 
        if isinstance(session_authuser, dict):
 
            try:
 
                return AuthUser.from_cookie(session_authuser)
 
            except UserCreationError as e:
 
                # container auth or other auth functions that create users on
 
                # the fly can throw UserCreationError to signal issues with
 
                # user creation. Explanation should be provided in the
 
                # exception object.
 
                from kallithea.lib import helpers as h
 
                h.flash(e, 'error', logf=log.error)
 

	
 
        # Authenticate by auth_container plugin (if enabled)
 
        if any(
 
            auth_modules.importplugin(name).is_container_auth
 
            for name in Setting.get_auth_plugins()
 
        ):
 
            try:
 
                user_info = auth_modules.authenticate('', '', request.environ)
 
            except UserCreationError as e:
kallithea/lib/middleware/simplehg.py
Show inline comments
 
@@ -75,234 +75,231 @@ class SimpleHg(BaseVCSController):
 
            return HTTPNotAcceptable('SSL REQUIRED !')(environ, start_response)
 

	
 
        ip_addr = self._get_ip_addr(environ)
 
        username = None
 
        # skip passing error to error controller
 
        environ['pylons.status_code_redirect'] = True
 

	
 
        #======================================================================
 
        # EXTRACT REPOSITORY NAME FROM ENV
 
        #======================================================================
 
        try:
 
            str_repo_name = environ['REPO_NAME'] = self.__get_repository(environ)
 
            assert isinstance(str_repo_name, str), str_repo_name
 
            repo_name = safe_unicode(str_repo_name)
 
            assert safe_str(repo_name) == str_repo_name, (str_repo_name, repo_name)
 
            log.debug('Extracted repo name is %s', repo_name)
 
        except Exception as e:
 
            log.error('error extracting repo_name: %r', e)
 
            return HTTPInternalServerError()(environ, start_response)
 

	
 
        # quick check if that dir exists...
 
        if not is_valid_repo(repo_name, self.basepath, 'hg'):
 
            return HTTPNotFound()(environ, start_response)
 

	
 
        #======================================================================
 
        # GET ACTION PULL or PUSH
 
        #======================================================================
 
        action = self.__get_action(environ)
 

	
 
        #======================================================================
 
        # CHECK ANONYMOUS PERMISSION
 
        #======================================================================
 
        if action in ['pull', 'push']:
 
            anonymous_user = self.__get_user('default')
 
            username = anonymous_user.username
 
            if anonymous_user.active:
 
                # ONLY check permissions if the user is activated
 
                anonymous_perm = self._check_permission(action, anonymous_user,
 
                                                        repo_name, ip_addr)
 
            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 this '
 
                              'repository as anonymous user')
 

	
 
                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 based authentication')
 
                pre_auth = auth_modules.authenticate('', '', environ)
 
                if pre_auth is not None and pre_auth.get('username'):
 
                    username = pre_auth['username']
 
                log.debug('PRE-AUTH got %s as username', username)
 

	
 
                # If not authenticated by the container, running basic auth
 
                if not username:
 
                    self.authenticate.realm = \
 
                        safe_str(self.config['realm'])
 
                    result = self.authenticate(environ)
 
                    if isinstance(result, str):
 
                        AUTH_TYPE.update(environ, 'basic')
 
                        REMOTE_USER.update(environ, result)
 
                        username = result
 
                    else:
 
                        return result.wsgi_application(environ, start_response)
 

	
 
                #==============================================================
 
                # CHECK PERMISSIONS FOR THIS REQUEST USING GIVEN USERNAME
 
                #==============================================================
 
                try:
 
                    user = self.__get_user(username)
 
                    if user is None or not user.active:
 
                        return HTTPForbidden()(environ, start_response)
 
                    username = user.username
 
                except Exception:
 
                    log.error(traceback.format_exc())
 
                    return HTTPInternalServerError()(environ, start_response)
 

	
 
                #check permissions for this repository
 
                perm = self._check_permission(action, user, repo_name, ip_addr)
 
                if not perm:
 
                    return HTTPForbidden()(environ, start_response)
 

	
 
        # extras are injected into mercurial UI object and later available
 
        # in hg hooks executed by kallithea
 
        from kallithea import CONFIG
 
        server_url = get_server_url(environ)
 
        extras = {
 
            'ip': ip_addr,
 
            'username': username,
 
            'action': action,
 
            'repository': repo_name,
 
            'scm': 'hg',
 
            'config': CONFIG['__file__'],
 
            'server_url': server_url,
 
            'make_lock': None,
 
            'locked_by': [None, None]
 
        }
 
        #======================================================================
 
        # MERCURIAL REQUEST HANDLING
 
        #======================================================================
 
        repo_path = os.path.join(safe_str(self.basepath), str_repo_name)
 
        log.debug('Repository path is %s', repo_path)
 

	
 
        # CHECK LOCKING only if it's not ANONYMOUS USER
 
        if username != User.DEFAULT_USER:
 
            log.debug('Checking locking on repository')
 
            (make_lock,
 
             locked,
 
             locked_by) = self._check_locking_state(
 
                            environ=environ, action=action,
 
                            repo=repo_name, user_id=user.user_id
 
                       )
 
            # store the make_lock for later evaluation in hooks
 
            extras.update({'make_lock': make_lock,
 
                           'locked_by': locked_by})
 

	
 
        fix_PATH()
 
        log.debug('HOOKS extras is %s', extras)
 
        baseui = make_ui('db')
 
        self.__inject_extras(repo_path, baseui, extras)
 

	
 
        try:
 
            log.info('%s action on Mercurial repo "%s" by "%s" from %s',
 
                     action, str_repo_name, safe_str(username), ip_addr)
 
            app = self.__make_app(repo_path, baseui, extras)
 
            result = app(environ, start_response)
 
            if action == 'push':
 
                result = WSGIResultCloseCallback(result,
 
                    lambda: self._invalidate_cache(repo_name))
 
            return result
 
        except RepoError as e:
 
            if str(e).find('not found') != -1:
 
                return HTTPNotFound()(environ, start_response)
 
        except HTTPLockedRC as e:
 
            # Before Mercurial 3.6, lock exceptions were caught here
 
            log.debug('Locked, response %s: %s', e.code, e.title)
 
            return e(environ, start_response)
 
        except Exception:
 
            log.error(traceback.format_exc())
 
            return HTTPInternalServerError()(environ, start_response)
 

	
 
    def __make_app(self, repo_name, baseui, extras):
 
        """
 
        Make an wsgi application using hgweb, and inject generated baseui
 
        instance, additionally inject some extras into ui object
 
        """
 
        class HgWebWrapper(hgweb_mod.hgweb):
 
            # Work-around for Mercurial 3.6+ causing lock exceptions to be
 
            # thrown late
 
            def _runwsgi(self, req, repo):
 
                try:
 
                    return super(HgWebWrapper, self)._runwsgi(req, repo)
 
                except HTTPLockedRC as e:
 
                    log.debug('Locked, response %s: %s', e.code, e.title)
 
                    req.respond(e.status, 'text/plain')
 
                    return ''
 

	
 
        return HgWebWrapper(repo_name, name=repo_name, baseui=baseui)
 

	
 
    def __get_repository(self, environ):
 
        """
 
        Gets repository name out of PATH_INFO header
 

	
 
        :param environ: environ where PATH_INFO is stored
 
        """
 
        try:
 
            environ['PATH_INFO'] = self._get_by_id(environ['PATH_INFO'])
 
            repo_name = '/'.join(environ['PATH_INFO'].split('/')[1:])
 
            if repo_name.endswith('/'):
 
                repo_name = repo_name.rstrip('/')
 
        except Exception:
 
            log.error(traceback.format_exc())
 
            raise
 

	
 
        return repo_name
 

	
 
    def __get_user(self, username):
 
        return User.get_by_username(username)
 

	
 
    def __get_action(self, environ):
 
        """
 
        Maps mercurial request commands into a clone,pull or push command.
 
        This should always return a valid command string
 
        Maps Mercurial request commands into 'pull' or 'push'.
 

	
 
        :param environ:
 
        """
 
        mapping = {'changegroup': 'pull',
 
                   'changegroupsubset': 'pull',
 
                   'stream_out': 'pull',
 
                   'listkeys': 'pull',
 
                   'unbundle': 'push',
 
                   'pushkey': 'push', }
 
        for qry in environ['QUERY_STRING'].split('&'):
 
            if qry.startswith('cmd'):
 
                cmd = qry.split('=')[-1]
 
                if cmd in mapping:
 
                    return mapping[cmd]
 

	
 
                return 'pull'
 
            parts = qry.split('=', 1)
 
            if len(parts) == 2 and parts[0] == 'cmd':
 
                cmd = parts[1]
 
                return mapping.get(cmd, 'pull')
 

	
 
        raise Exception('Unable to detect pull/push action !!'
 
                        'Are you using non standard command or client ?')
 

	
 
    def __inject_extras(self, repo_path, baseui, extras={}):
 
        """
 
        Injects some extra params into baseui instance
 

	
 
        also overwrites global settings with those takes from local hgrc file
 

	
 
        :param baseui: baseui instance
 
        :param extras: dict with extra params to put into baseui
 
        """
 

	
 
        hgrc = os.path.join(repo_path, '.hg', 'hgrc')
 

	
 
        repoui = make_ui('file', hgrc, False)
 

	
 
        if repoui:
 
            #overwrite our ui instance with the section from hgrc file
 
            for section in ui_sections:
 
                for k, v in repoui.configitems(section):
 
                    baseui.setconfig(section, k, v)
 
        _set_extras(extras)
0 comments (0 inline, 0 general)