Re: PostmasterContext survives into parallel workers!?

On Mon, Aug 1, 2016 at 7:42 PM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> Thomas Munro <thomas(dot)munro(at)enterprisedb(dot)com> writes:
>> I found this apparently unresolved bug report about glibc fork()
>> inside a signal handler deadlocking:
>> https://sourceware.org/bugzilla/show_bug.cgi?id=4737
>
>> I wonder if that could bite postmaster.
>
> I seriously doubt it. The key thing about the postmaster is that
> it runs with signals blocked practically everywhere. So we're not
> taking risks of a signal handler interrupting, say, malloc()
> (which seemed to be the core of at least the first example given
> in that report). This is what makes me dubious that getting rid
> of doing work in the postmaster's signal handlers is really going
> to add any noticeable increment of safety. It might make the
> code look cleaner, but I'm afraid it's going to be a lot of churn
> for not much gain.

It's not just a cosmetic issue.

See https://www.postgresql.org/message-id/CA+Tgmobr+Q2WgWeasdbDNefVwJkAGALxA=-VtEGNtQgL1V2Ryw@mail.gmail.com
and d0410d66037c2f3f9bee45e0a2db9e47eeba2bb4.

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