Re: Reviewing freeze map code

On Wed, Jun 15, 2016 at 6:26 PM, Robert Haas <robertmhaas@gmail.com> wrote:
>
> On Wed, Jun 15, 2016 at 2:41 AM, Thomas Munro
> <thomas.munro@enterprisedb.com> wrote:
> > I spent some time chasing down the exact circumstances.  I suspect
> > that there may be an interlocking problem in heap_update.  Using the
> > line numbers from cae1c788 [1], I see the following interaction
> > between the VACUUM, UPDATE and SELECT (pg_check_visible) backends, all
> > in reference to the same block number:
> >
> >   [VACUUM] sets all visible bit
> >
> >   [UPDATE] heapam.c:3931 HeapTupleHeaderSetXmax(oldtup.t_data, xmax_old_tuple);
> >   [UPDATE] heapam.c:3938 LockBuffer(buffer, BUFFER_LOCK_UNLOCK);
> >
> >   [SELECT] LockBuffer(buffer, BUFFER_LOCK_SHARE);
> >   [SELECT] observes VM_ALL_VISIBLE as true
> >   [SELECT] observes tuple in HEAPTUPLE_DELETE_IN_PROGRESS state
> >   [SELECT] barfs
> >
> >   [UPDATE] heapam.c:4116 visibilitymap_clear(...)
>
> Yikes: heap_update() sets the tuple's XMAX, CMAX, infomask, infomask2,
> and CTID without logging anything or clearing the all-visible flag and
> then releases the lock on the heap page to go do some more work that
> might even ERROR out.