Re: why local_preload_libraries does require a separate directory ? - Mailing list pgsql-hackers

From Tomas Vondra
Subject Re: why local_preload_libraries does require a separate directory ?
Date
Msg-id 4EDBEC84.2030809@fuzzy.cz
Whole thread Raw
In response to Re: why local_preload_libraries does require a separate directory ?  (Tom Lane <tgl@sss.pgh.pa.us>)
Responses Re: why local_preload_libraries does require a separate directory ?  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-hackers
On 4.12.2011 22:16, Tom Lane wrote:
> Tomas Vondra <tv@fuzzy.cz> writes:
>> On 3.12.2011 18:53, Tom Lane wrote:
>>> Security: it lets the DBA constrain which libraries are loadable this way.
> 
>> But local_preload_libraries can be set only in postgresql.conf, right?
> 
> No.  It's PGC_BACKEND, which means it can be set at connection start by
> a client (eg, via PGOPTIONS).
> 
>> The problem I'm trying to solve right now is that I do have an extension
>> that needs to install two .so libraries - one loaded using
>> shared_preload_libraries, one loaded using local_preload_libraries.
> 
> Um ... why would you design it like that?

Well, the purpose of the extension is to limit number of connections by
db/user/IP. It's using pg_stat_activity and auth hook, and the rules are
loaded from a file into a shared memory segment.

So I need to:

1) allocate memory to keep the rules (so it has to be loaded using  shared_preload_libraries)

2) check the rules (this is done in auth hook) using pg_stat_activity

The backends are added to pg_stat_activity after the auth hook finishes,
which means possible race conditions (backends executed at about the
same time don't see each other in pg_stat_activity). So I use an
exclusive lock that's acquired before reading pg_stat_activity and
released after the pg_stat_activity is updated.

That's the only thing the library loaded using local_preload_libraries
does - it releases the lock.

Tomas


pgsql-hackers by date:

Previous
From: Pavel Stehule
Date:
Subject: Re: planner fails on HEAD
Next
From: Martin Pihlak
Date:
Subject: Re: logging in high performance systems.