Re: Regarding #8580 - Mailing list pgadmin-hackers

From Dave Page
Subject Re: Regarding #8580
Date
Msg-id CA+OCxozpzppfn9hJky9FvvfG89=r+ZV7Ui85H+ntESAcG0i8og@mail.gmail.com
Whole thread Raw
In response to Re: Regarding #8580  (Khushboo Vashi <khushboo.vashi@enterprisedb.com>)
Responses Re: Regarding #8580
List pgadmin-hackers


On Fri, 9 May 2025 at 12:48, Akshay Joshi <akshay.joshi@enterprisedb.com> wrote:


On Fri, May 9, 2025 at 4:50 PM Dave Page <dpage@pgadmin.org> wrote:


On Fri, 9 May 2025 at 12:14, Akshay Joshi <akshay.joshi@enterprisedb.com> wrote:


On Fri, May 9, 2025 at 4:20 PM Dave Page <dpage@pgadmin.org> wrote:


On Fri, 9 May 2025 at 11:34, Khushboo Vashi <khushboo.vashi@enterprisedb.com> wrote:


On Fri, May 9, 2025 at 3:23 PM Dave Page <dpage@pgadmin.org> wrote:
Hi

On Fri, 9 May 2025 at 08:45, Akshay Joshi <akshay.joshi@enterprisedb.com> wrote:
Hi Hackers/Dave,

I have started working on issue #8580, where the correct error message should be displayed based on the user's authentication source when an incorrect password is provided.

Actual Issue: The admin has configured AUTHENTICATION_SOURCES = ['internal', 'ldap']. A user with the email a@xyz.com exists only as an internal user in the database, and there is no corresponding LDAP entry for this user. When this user attempts to log in with an incorrect password, the system first tries internal authentication, which fails. It then proceeds to check the next authentication source (LDAP), as per the configured logic. Since no matching LDAP user exists, an LDAP-related error is returned, even though the user is intended to be authenticated only internally. His/her account will never get locked.

This behavior appears to be incorrect to me. I’m proposing two possible solutions to address it:
Solution 1 (Logic Changes): 
Scenario 1: ['internal', 'ldap']:
  • If a user exists in the database with the specified authentication source (internal), attempt to authenticate using internal. If authentication fails, return an error. No need to check for the LDAP or next auth source.
Yes. 
  • If no user-auth source combination is found for internal, proceed to the next authentication source (LDAP). Attempt LDAP login, and if successful (and auto-create is enabled), create the user in the database.
Yes.
 
Scenario 2: ['ldap', 'internal'] 
  • If the LDAP user does not exist in the database, but the same user exists as an internal user, first try LDAP authentication. If it fails, fall back to internal or the next configured auth source in the list. 
Yes. 
  • If the LDAP user does exist in the database, attempt to authenticate via LDAP. If LDAP authentication fails, return the error without checking for the next authentication source.
Yes.


If the user is registered for multiple authentications (per entries in our database), the next in line should be checked if one fails.=

I think that's reasonable, but *only* in that case where there's another source already present in the DB.

  In that case, the core issue remains unresolved. As mentioned earlier, the system checks internal authentication first (based on the configured order), and then attempts LDAP if the user exists. However, it consistently returns an error for LDAP, and the account is never locked even after exceeding the maximum number of login attempts.

So, just lock all matching accounts. 

    One more scenario I just thought of with 'Auto Create User'. Suppose that a@xyz.com exists as an internal user in the database, but there is no corresponding LDAP entry. When the user attempts to log in with an incorrect password, the system checks the database for entries from other authentication sources and throws an error. In this case, the 'Auto Create User' functionality will not be triggered.

    I think we can either keep the current behavior as it is and close the issue (since I reported it), or add a rule with documentation saying that login should follow the order of the authentication sources. In most cases, users who prefer LDAP can just set the auth source to ['ldap', 'internal'], which should work fine. 

Right, I believe that would be the typical case.
 
--

pgadmin-hackers by date:

Previous
From: Dave Page
Date:
Subject: Re: Regarding #8580
Next
From: Akshay Joshi
Date:
Subject: Re: Regarding #8580