On 2014-03-13 11:11:51 -0400, Tom Lane wrote:
> Andres Freund <andres@2ndquadrant.com> writes:
> > On 2014-03-13 10:26:11 -0400, Tom Lane wrote:
> >> No, because relcache doesn't store security labels to start with.
> >> There's a separate catalog cache for security labels, I believe,
> >> and invalidating entries in that ought to be sufficient.
>
> > There doesn't seem to be any form of system managed cache for security
> > labels afaics. Every lookup does a index scan. I currently don't see how
> > I could build a cache in userland that'd invalidate if either a) the
> > underlying object changes b) the label changes.
>
> If there's not a catcache for pg_seclabels, I'd have no objection
> to adding one.
Ok. That's an easy enough patch, would anyone object to adding that now?
> As for your "userland cache" objection, you certainly
> could build such a thing using the existing inval callbacks (if we
> had a catcache on pg_seclabels), and in any case what have userland
> caches got to do with relcache?
I don't think I've said anything about relcaches being required for
this. It came up in 20140313132604.GG8268@awork2.anarazel.de, but that
was because we were just talking table level there, and it's a tad
easier to hook into relcache invalidation callbacks than catcache ones.
That said, for a relation level cache that refer's to the table's
definition, you really *do* need a relcache invalidation callback, not
just a catcache callback. There's a fair number of places that do a
CacheInvalidateRelcache() to trigger invals.
Greetings,
Andres Freund
-- Andres Freund http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training &
Services