Horton

Re: ExecRTCheckPerms() and many prunable partitions (checkAsUser)

Justin Pryzby <pryzby@telsasoft.com>

View thread View on postgr.es
From: Justin Pryzby <pryzby@telsasoft.com>
To: Amit Langote <amitlangote09@gmail.com>
Cc: Alvaro Herrera <alvherre@alvh.no-ip.org>, Ian Lawrence Barwick <barwick@gmail.com>, Tom Lane <tgl@sss.pgh.pa.us>, David Rowley <dgrowleyml@gmail.com>, Greg Stark <stark@mit.edu>, Julien Rouhaud <rjuju123@gmail.com>, pgsql-hackers@lists.postgresql.org
Date: 2022-12-11T14:21:58Z
Lists: pgsql-hackers 
On Sun, Dec 11, 2022 at 06:25:48PM +0900, Amit Langote wrote:
> On Sun, Dec 11, 2022 at 5:17 AM Justin Pryzby <pryzby@telsasoft.com> wrote:
> > The original code rechecks rte->checkAsUser with the rte of the parent
> > rel.  The patch changed to access onerel instead, but that's not updated
> > after looping to find the parent.
> >
> > Is that okay ?  It doesn't seem intentional, since "userid" is still
> > being recomputed, but based on onerel, which hasn't changed.  The
> > original intent (since 553d2ec27) is to recheck the parent's
> > "checkAsUser".
> >
> > It seems like this would matter for partitioned tables, when the
> > partition isn't readable, but its parent is, and accessed via a view
> > owned by another user?
> 
> Thanks for pointing this out.
> 
> I think these blocks which are rewriting userid to basically the same
> value should have been removed from these sites as part of 599b33b94.

I thought maybe; thanks for checking.

Little nitpicks:

001:
Fine to use the same userid as it's same in all
=> the same

002:
give that it's a subquery rel.
=> given

-- 
Justin

Commits

  1. Add a test case for a316a3bc

  2. Correctly set userid of subquery relations' child rels

  3. Fix buggy recursion in flatten_rtes_walker().

  4. Remove some dead code in selfuncs.c

  5. Update outdated comment in ApplyRetrieveRule

  6. Rework query relation permission checking

  7. Generalize ri_RootToPartitionMap to use for non-partition children

  8. Stop accessing checkAsUser via RTE in some cases

  9. Add 'missing_ok' argument to build_attrmap_by_name