Feature #1913
closedLDAP - authenticate as user
70%
Description
The attached patch allows to bind to the ldap server as the user logging in (instead of either anonymous or specific admin account).
The idea behind the patch is quite simple: When configuring an LDAP source one may or may not specify an "Account". If no account is specified, Redmine will bind anonymously. If the account is specified Redmine binds as that user.
The patch introduces a third state: bind as user. When the account is specified as (for example) "uid=$login,ou=people,dc=example,dc=com" $login is replaced with the login name and the password given by the user is used. Therefor there is no need to have an LDAP directory that is readable by anonymous bind and there is no need to have a password saved into the database.
I tried to make the patch as unintrusive as possible while completely being compatible to the way things are now.
Files
Related issues
Updated by Markus Peter about 16 years ago
I applied this patch on a development checkout (revision 1901) and it works fine!
It took me a while to figure out the ldap settings.
In the Account field, I entered[domain]\$login
and in Base DNCN=users,DC=people,DC=example,DC=com
This finally worked for me, no password stored anywhere.
Updated by Martin Bächtold about 16 years ago
This patch also worked for me, thank you very much.
Updated by Jérémie Delaitre almost 16 years ago
Will this patch be included into the core ?
Updated by Eric Davis almost 16 years ago
Could someone update the patch to include a few test cases?
Updated by Daniel Marczisovszky over 15 years ago
I've created a patch for alias dereferencing (#3253) and I included your patch. It also adds options for custom search filtering that was influenced by your patch, it uses the same syntax. Could you please try it out?
Updated by Adi Kriegisch over 15 years ago
To make this patch usable with Apache authentication against Redmine for SVN access (ie. all the stuff that is located in redmine/extra/svn) it is necessary to patch Redmine.pm to as well understand how to treat $login in config.
btw. Arnaud Martel did a great job in enhancing Redmine.pm. See #3712.
Updated by Eric Davis almost 15 years ago
- Category changed from Accounts / authentication to LDAP
Can someone post a simple OpenLDAP configuration that removes the anonymous binding so I can test this? I've been working in the LDAP code recently and would like to apply this.
Updated by Adi Kriegisch almost 15 years ago
Eric Davis wrote:
Can someone post a simple OpenLDAP configuration that removes the anonymous binding so I can test this? I've been working in the LDAP code recently and would like to apply this.
Cool! Looking forward to you patch!
Assuming you've got a working LDAP server adding an access control list is the only thing required (if you already have some watch out for the correct sorting -- they are processed top to bottom and first match stops further processing).access to attrs=userPassword by dn.base="uid=root,dc=mydomain,dc=com" write
_ _ by self write
_ _ by anonymous auth
(_ _ == just spaces)
should be quite self explaining: in contrary to SQL-queries LDAP just hides attributes if you're not allowed to see them ("read" permission) so you will still be able to get a list of all users. Just the "userPassword" field is hidden. "anonymous" connections may just use this attribute to authenticate whereas the authenticated user himself may read and change his password and root -- as usual -- may do anything!
In real life you might add other ACL as well and you might want to add the "ssf" parameter as well (security strenght factor that specifies if SSL/TLS is required for access to a certain attribute).
Hope this helps! Feel free to contact me if you need any further assistance!
Updated by Felix Schäfer almost 15 years ago
This is a "light" version of what we have, so it's kinda untested in this form, but it should give read or read/write permission to some attributes only to some users.
to attrs=userPassword,shadowLastChange by self write by anonymous auth by * none to dn.one="ou=People,dc=example,dc=com" attrs=uid # This assumes a flat user directory, change one to children to extend it to any child of ou=People at any depth by self read by users read # this ensures that any user in the LDAP can search/read for uids to enable the search for a cn based on the uid, redmine obviously needs a "user" somewhere in the LDAP tree for that to work by * none to dn.children="ou=People,dc=example,dc=com" attrs=givenName,sn,mail # We have more attributes here, but that should be the only three redmine needs atm by self write by users read # same comment as above by * none to dn.base="" by * read to * by users read # not sure how much this part is needed, as it gives a blanket read access to all users in the LDAP, but should be ok for testing the need for user credentials to read the LDAP by * none
I also removed all our "extra" admin accesses and stuff, catch me on IRC if you need more info/help with that.
Updated by Felix Schäfer almost 15 years ago
By the way, you won't need a "full" user for the redmine to access the LDAP, "only" an LDAP object with a password, our "users" for app access export to something like that:
dn: cn=Redmine,ou=Apps,ou=System,dc=example,dc=com cn: Redmine objectClass: namedObject objectClass: top objectClass: simpleSecurityObject userPassword: {SSHA}somefunnyhash
Updated by Antoine Beaupré almost 15 years ago
- File Redmine-app-models-auth_source_ldap-0.9.1-2.diff Redmine-app-models-auth_source_ldap-0.9.1-2.diff added
- % Done changed from 0 to 50
So here's a patch that applies to the Debian 0.9.1 package that improves a bit on what's already here in that it avoids doing a second bind if we're using the introduced "bind as user" setting.
Updated by Antoine Beaupré almost 15 years ago
- File 1913_redmine_bind_as_user.diff 1913_redmine_bind_as_user.diff added
- % Done changed from 50 to 70
... and here's an untested patch for current head. it refactors get_user_dn() to use an internal ldap_con to avoid binding twice with the ldap server when using user-level bindings.
i think the UI could still need some love to explain the $login hack, but in the meantime this makes more sense in the LDAP world...
Updated by Bernhard Furtmueller almost 15 years ago
I had the very same problem, tried to search but didn´t find this solution in the past.
Therefore a did the "reinvent the wheel" approach and came up with my patch
which introduces a new domain field.
See forum entry here:
http://www.redmine.org/boards/1/topics/11119
And the patch (which works at least against active directory):
http://www.redmine.org/attachments/3176/patch0.patch
I´ll try the patch (is it supposed to work against active directory?)
attached to this tracker asap and will report.
br,
bernhard
Updated by Markus Peter over 14 years ago
- File Bind_as_user_LDAP.diff Bind_as_user_LDAP.diff added
Antoine Beaupré wrote:
... and here's an untested patch for current head. it refactors get_user_dn() to use an internal ldap_con to avoid binding twice with the ldap server when using user-level bindings.
The patch did throw an error:
NoMethodError (undefined method `ldap_con=' for #<AuthSourceLdap:0x8c22f74>): app/models/auth_source_ldap.rb:38:in `authenticate' app/models/user.rb:105:in `try_to_login' app/controllers/account_controller.rb:147:in `password_authentication' app/controllers/account_controller.rb:142:in `authenticate_user' app/controllers/account_controller.rb:30:in `login'
Chances are I messed up something...
I included a patch which works with the current head, though it still calls initialize_ldap_con
twice, once in authenticate
and once in authenticate_dn
.
Updated by Felix Schäfer over 14 years ago
Markus Peter wrote:
I included a patch which works with the current head, though it still calls
initialize_ldap_con
twice, once inauthenticate
and once inauthenticate_dn
.
There are environments in which you might need to first bind as the "redmine user" to the LDAP to determine the DN corresponding to a certain login as not all LDAP setups support searching for login as anonymous.
Updated by Markus Peter over 14 years ago
Felix Schäfer wrote:
There are environments in which you might need to first bind as the "redmine user" to the LDAP to determine the DN corresponding to a certain login as not all LDAP setups support searching for login as anonymous.
Our AD does not support serching for logins as anonymous, that's why we use this patch to connect with the login/password supplied by the user and the parameters provided in the LDAP configuration without having to enter a user/pwd for each LDAP config.
Not sure whether this works with self-registration, though.
Antoine's patch is certainly better, but unless I messed up something there may be missing something to make it work.
Updated by Bernhard Furtmueller over 14 years ago
Had the same problem, so I´ll post my approach here. This is against trunk @3625
This adds a domain field which will be prefixed to the userid.
It also allows self-registration.
HTH,
bernhard
commit c6b87839849899fb2c24fde1533224f60818074e Author: Bernhard Furtmueller <bernhard.furtmueller@hilti.com> Date: Tue Mar 30 13:37:14 2010 +0000 adding a domain field in order to allow direct active directory authentication without requiring a read only ads user. forward port of 0b9ee54dafe21140bf694bf968431633d4ec09b5 only with lang en and de diff --git a/app/models/auth_source_ldap.rb b/app/models/auth_source_ldap.rb index d2a7e70..a7bb7ba 100644 --- a/app/models/auth_source_ldap.rb +++ b/app/models/auth_source_ldap.rb @@ -21,7 +21,7 @@ require 'iconv' class AuthSourceLdap < AuthSource validates_presence_of :host, :port, :attr_login validates_length_of :name, :host, :account_password, :maximum => 60, :allow_nil => true - validates_length_of :account, :base_dn, :maximum => 255, :allow_nil => true + validates_length_of :account, :domain, :base_dn, :maximum => 255, :allow_nil => true validates_length_of :attr_login, :attr_firstname, :attr_lastname, :attr_mail, :maximum => 30, :allow_nil => true validates_numericality_of :port, :only_integer => true @@ -33,7 +33,7 @@ class AuthSourceLdap < AuthSource def authenticate(login, password) return nil if login.blank? || password.blank? - attrs = get_user_dn(login) + attrs = get_user_dn(login,password) if attrs && attrs[:dn] && authenticate_dn(attrs[:dn], password) logger.debug "Authentication successful for '#{login}'" if logger && logger.debug? @@ -100,8 +100,10 @@ class AuthSourceLdap < AuthSource end # Get the user's dn and any attributes for them, given their login - def get_user_dn(login) - ldap_con = initialize_ldap_con(self.account, self.account_password) + def get_user_dn(login,password) + #ldap_con = initialize_ldap_con(self.account, self.account_password) + domain.blank? ? ldap_con = initialize_ldap_con(self.account, self.account_password) : ldap_con = initialize_ldap_con(domain + "\\" + login, password); + login_filter = Net::LDAP::Filter.eq( self.attr_login, login ) object_filter = Net::LDAP::Filter.eq( "objectClass", "*" ) attrs = {} diff --git a/app/views/auth_sources/_form.rhtml b/app/views/auth_sources/_form.rhtml index 9ffffaf..f023bce 100644 --- a/app/views/auth_sources/_form.rhtml +++ b/app/views/auth_sources/_form.rhtml @@ -11,6 +11,9 @@ <p><label for="auth_source_port"><%=l(:field_port)%> <span class="required">*</span></label> <%= text_field 'auth_source', 'port', :size => 6 %> <%= check_box 'auth_source', 'tls' %> LDAPS</p> +<p><label for="auth_source_domain"><%=l(:field_domain)%></label> +<%= text_field 'auth_source', 'domain' %></p> + <p><label for="auth_source_account"><%=l(:field_account)%></label> <%= text_field 'auth_source', 'account' %></p> diff --git a/config/locales/de.yml b/config/locales/de.yml index 982452e..d79c20a 100644 --- a/config/locales/de.yml +++ b/config/locales/de.yml @@ -268,6 +268,7 @@ de: field_port: Port field_account: Konto field_base_dn: Base DN + field_domain: Domäne field_attr_login: Mitgliedsname-Attribut field_attr_firstname: Vorname-Attribut field_attr_lastname: Name-Attribut diff --git a/config/locales/en.yml b/config/locales/en.yml index 4082670..b155582 100644 --- a/config/locales/en.yml +++ b/config/locales/en.yml @@ -243,6 +243,7 @@ en: field_port: Port field_account: Account field_base_dn: Base DN + field_domain: Domain field_attr_login: Login attribute field_attr_firstname: Firstname attribute field_attr_lastname: Lastname attribute diff --git a/db/migrate/20100330124427_add_auth_sources_domain.rb b/db/migrate/20100330124427_add_auth_sources_domain.rb new file mode 100644 index 0000000..f9d1de5 --- /dev/null +++ b/db/migrate/20100330124427_add_auth_sources_domain.rb @@ -0,0 +1,9 @@ +class AddAuthSourcesDomain < ActiveRecord::Migration + def self.up + add_column :auth_sources, :domain, :string, :default => 'none', :null => false + end + + def self.down + remove_column :auth_sources, :domain + end +end
Updated by Antoine Beaupré almost 14 years ago
For me the patch in comment #17 was unclear: first, there's commented code in there and second, I don't understand the reason for the 'domain' field. The patch I am providing here is just a port of the one in comment #14, which is itself a port of my patch, which was a port of... well, you see where I'm going. :)
Basically, I believe that using $login in the field is (a) much more powerful and (b) flexible enough to cover for the "domain" case, which is really unclear to me what it does.
Can we get this committed please? This issue has been opened for 2 years and a patch has been available for over a year now. I've been using it in production here for the last 10 months without any problems.
The patch applies to 1.0.1 but it should be trivial to port it to trunk. Besides, i did that earlier and that didn't seem to favor inclusion so I'll just try to port this to new releases as we upgrade, but I'd really like to see this hit the trunk so i don't have to maintain this silly patch.
Updated by Adi Kriegisch almost 14 years ago
Thanks, Antoine, for forward porting this patch. (Just a minor correction: this patch is available for two years now -- and it is still working in production ;-)
Probably main developers are lacking something? What do I/we need to do to finally get this included into Redmine?
Eric Davis (comment 4) asked for test cases. How should they look like? Is it enough to proof that this patch generates a correct login dn?
Probably documentation is missing? How should the documentation look like? The very first comment explains how to bind against AD. Comment 8 and 9 provide useful LDAP config snippets for OpenLDAP. Anything more needed?
I'd love to have this upstream: rebuilding packages all the time to get this working again and again is just annoying.
Updated by Daniel Ritz almost 13 years ago
+1
Even the minimal patch in 1913_redmine_bind_as_user2.diff is a big improvement. Using that one at work to authenticate against a Windows AD without requiring an extra user (which is not easy to get...corporate politics).
Updated by Antoine Beaupré almost 13 years ago
I can confirm this is still working in production, and I painstakingly update this at every release. So far there was no change in the 1.1 release, but we'll see for 1.3.
How do we get patches merged into redmine? I have a good lot waiting in the queue and they seem to be getting no attention whatsoever...
Updated by Antoine Beaupré over 12 years ago
Hello? Anyone?
Should I submit this to chiliproject instead?
Updated by Jean-Philippe Lang over 12 years ago
Would you be able to add a test case to the patch ? Or should I take care of it ?
Updated by Adi Kriegisch over 12 years ago
Jean-Philippe, it would be great if you can add test cases. Please let me know if you lack anything for adding this patch. I'd gladly help!
Thank you for considering this patch!
Updated by Antoine Beaupré over 12 years ago
I am sorry I am not familiar enough with Redmine's unit testing to provide a test case here. Besides, I think it would require a running LDAP server, which is not trivial...
It would be awesome if someone else could provide that test case though. Thanks for looking into this patch!
Updated by Jean-Philippe Lang over 12 years ago
- Status changed from New to Closed
- Assignee set to Jean-Philippe Lang
- Target version set to 1.4.0
Feature added in r9241 - r9243 with slight changes and tests. The initial patch was breaking 2 tests and i think it's safer to escape the submitted login.
Antoine Beaupré wrote:
Besides, I think it would require a running LDAP server, which is not trivial...
This was already required to run the full test suite, so adding tests for this new feature was pretty straightforward.
Thanks for your contribution.
Updated by Harley Laue over 12 years ago
extra/svn/Redmine.pm still needs patched (which was supplied) to work correctly. Without this patch, SVN read and/or write is completely broken while using this feature within Redmine.
Updated by Andrew Kohlsmith over 10 years ago
There is a more generic issue with binding to the AD. I ran into this and don't have a good solution, but I'll describe it here so that hopefully others will find it.
Windows 2003 seems to confuse the notion of binding to the LDAP server and logging in to a workstation. If you have a domain user who is restricted to logging in only from a specific workstation or workstations, the bind as that user will fail because AD sees the bind, validates the password and then notices that the AD server itself is not in the list of allowed workstations. The specific LDAP error is
W80090308: LdapErr: DSID-0C090334, comment: AcceptSecurityContext error, data 531, vece
where the 531
corresponds to "User not allowed to logon at this computer".
I don't know enough about AD to know if there is a way to say "any valid account can bind the AD server" while maintaining the workstation login restriction. The only way around it that I have found is to add the NetBIOS name of the AD server to the restricted user's allowed workstation list. This is clearly not a great solution.