Horton

Re: SIGQUIT handling, redux

Tom Lane <tgl@sss.pgh.pa.us>

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Andres Freund <andres@anarazel.de>
Cc: pgsql-hackers@lists.postgresql.org
Date: 2020-09-09T23:24:55Z
Lists: pgsql-hackers

Attachments

Here's a draft patch that I think would be reasonable to back-patch.
(Before v13, we'd need a bespoke SIGQUIT handler to substitute for
SignalHandlerForCrashExit, but that's easy enough.)

Aside from comment updates, this

* uses SignalHandlerForCrashExit for SIGQUIT

* renames startup_die per your request

* moves BackendInitialize's interrupt-re-disabling code up a bit to
reduce the scope where these interrupts are active.  This doesn't
make things a whole lot safer, but it can't hurt.

I'll take a closer look at the idea of using _exit(1) tomorrow,
but I'd be pretty hesitant to back-patch that.

			regards, tom lane

Commits

  1. Log a message when resorting to SIGKILL during shutdown/crash recovery.

  2. Don't run atexit callbacks during signal exits from ProcessStartupPacket.

  3. Use _exit(2) for SIGQUIT during ProcessStartupPacket, too.

  4. Make archiver's SIGQUIT handler exit via _exit().