Thread: Re: [COMMITTERS] pgsql: Make it easy to detach completely from shared memory.

On 18 March 2014 11:59, Robert Haas <rhaas@postgresql.org> wrote:
> Make it easy to detach completely from shared memory.
>
> The new function dsm_detach_all() can be used either by postmaster
> children that don't wish to take any risk of accidentally corrupting
> shared memory; or by forked children of regular backends with
> the same need.  This patch also updates the postmaster children that
> already do PGSharedMemoryDetach() to do dsm_detach_all() as well.
>
> Per discussion with Tom Lane.

I think we need to document exactly why dsm_detach_all() isn't simply
part of PGSharedMemoryDetach() ?

Having two calls seems like a recipe for error in core and extensions.

Perhaps we should consider a parameter for PGSharedMemoryDetach() ?

-- Simon Riggs                   http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training & Services

On Tue, Mar 18, 2014 at 8:41 AM, Simon Riggs <simon@2ndquadrant.com> wrote:
> On 18 March 2014 11:59, Robert Haas <rhaas@postgresql.org> wrote:
>> Make it easy to detach completely from shared memory.
>>
>> The new function dsm_detach_all() can be used either by postmaster
>> children that don't wish to take any risk of accidentally corrupting
>> shared memory; or by forked children of regular backends with
>> the same need.  This patch also updates the postmaster children that
>> already do PGSharedMemoryDetach() to do dsm_detach_all() as well.
>>
>> Per discussion with Tom Lane.
>
> I think we need to document exactly why dsm_detach_all() isn't simply
> part of PGSharedMemoryDetach() ?
>
> Having two calls seems like a recipe for error in core and extensions.
>
> Perhaps we should consider a parameter for PGSharedMemoryDetach() ?

Yeah, maybe.  It seems like a possible modularity violation, because
the PGSharedMemory... stuff has heretofore not needed to know anything
about DSM, and apart from this one function, it still wouldn't.  On
the other hand, your argument is good, too.

-- 
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company

Robert Haas <robertmhaas@gmail.com> writes:
> On Tue, Mar 18, 2014 at 8:41 AM, Simon Riggs <simon@2ndquadrant.com> wrote:
>> Perhaps we should consider a parameter for PGSharedMemoryDetach() ?

> Yeah, maybe.  It seems like a possible modularity violation, because
> the PGSharedMemory... stuff has heretofore not needed to know anything
> about DSM, and apart from this one function, it still wouldn't.

That was exactly the reason we rejected that design upthread.
PGSharedMemoryDetach is specific to the main shmem segment, and in fact
has multiple OS-dependent implementations.

You could make an argument for inventing some new wrapper function that
calls both PGSharedMemoryDetach and dsm_detach_all, but I don't believe
that the existing flavors of that function should know about DSM.
        regards, tom lane

On 18 March 2014 13:51, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Robert Haas <robertmhaas@gmail.com> writes:
>> On Tue, Mar 18, 2014 at 8:41 AM, Simon Riggs <simon@2ndquadrant.com> wrote:
>>> Perhaps we should consider a parameter for PGSharedMemoryDetach() ?
>
>> Yeah, maybe.  It seems like a possible modularity violation, because
>> the PGSharedMemory... stuff has heretofore not needed to know anything
>> about DSM, and apart from this one function, it still wouldn't.
>
> That was exactly the reason we rejected that design upthread.
> PGSharedMemoryDetach is specific to the main shmem segment, and in fact
> has multiple OS-dependent implementations.
>
> You could make an argument for inventing some new wrapper function that
> calls both PGSharedMemoryDetach and dsm_detach_all, but I don't believe
> that the existing flavors of that function should know about DSM.

I'm not bothered which we choose, as long as its well documented to
ensure people that use those calls don't detach from just one when
they really would wish to detach from both.

-- Simon Riggs                   http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training & Services