Re: Proposal: Save user's original authenticated identity for logging - Mailing list pgsql-hackers

From Michael Paquier
Subject Re: Proposal: Save user's original authenticated identity for logging
Date
Msg-id YGqkZeicvgHbMVvO@paquier.xyz
Whole thread Raw
In response to Re: Proposal: Save user's original authenticated identity for logging  (Michael Paquier <michael@paquier.xyz>)
Responses Re: Proposal: Save user's original authenticated identity for logging  (Jacob Champion <pchampion@vmware.com>)
List pgsql-hackers
On Sat, Apr 03, 2021 at 09:30:25PM +0900, Michael Paquier wrote:
> Slight rebase for this one to take care of the updates with the SSL
> error messages.

I have been looking again at that and applied it as c50624cd after
some slight modifications.  Attached is the main, refactored, patch
that plugs on top of the existing infrastructure.  connect_ok() and
connect_fails() gain two parameters each to match or to not match the
logs of the backend, with a truncation of the logs done before any
connection attempt.

I have spent more time reviewing the backend code while on it and
there was one thing that stood out:
+       ereport(FATAL,
+               (errmsg("connection was re-authenticated"),
+                errdetail_log("previous ID: \"%s\"; new ID: \"%s\"",
+                              port->authn_id, id)));
This message would not actually trigger because auth_failed() is the
code path in charge of showing an error here, so this could just be
replaced by an assertion on authn_id being NULL?  The contents of this
log were a bit in contradiction with the comments a couple of lines
above anyway.  Jacob, what do you think?
--
Michael

Attachment

pgsql-hackers by date:

Previous
From: Amul Sul
Date:
Subject: Re: [Patch] ALTER SYSTEM READ ONLY
Next
From: Bharath Rupireddy
Date:
Subject: Re: TRUNCATE on foreign table