Horton

Re: PostgreSQL crashes with SIGSEGV

Peter Geoghegan <pg@bowt.ie>

View thread View on postgr.es
From: Peter Geoghegan <pg@bowt.ie>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Aleksandr Parfenov <a.parfenov@postgrespro.ru>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Date: 2018-02-23T00:22:47Z
Lists: pgsql-bugs, pgsql-hackers

Attachments

On Mon, Feb 12, 2018 at 6:15 PM, Peter Geoghegan <pg@bowt.ie> wrote:
> * For 9.5 and 9.6, the approach taken in bugfix commit d8589946d
> should be taken even further -- we should always copy. Moreover, we
> should always copy within tuplesort_getdatum(), for the same reasons.
>
> * For 9.5, 9.6, 10, and master, we should make sure that
> tuplesort_getdatum() uses the caller's memory context. The fact that
> it doesn't already do so seems like a simple oversight. We should do
> this to be consistent with tuplesort_gettupleslot(). (This isn't
> critical, but seems like a good idea.)
>
> * For 9.5, 9.6, 10, and master, we should adjust some comments from
> tuplesort_getdatum() callers, so that they no longer say that
> tuplesort datum tuple memory lives in tuplesort context. That won't be
> true anymore.

Attached patches do it that way. I'm happy with what I came up with,
which is a lot simpler than my first approach. The extra copying seems
likely to be well worth it, since it is fairly isolated in practice,
especially on 9.6. There is no extra copying from v10+, since they
don't need the first fix at all.

-- 
Peter Geoghegan

Commits

  1. Fix actual and potential double-frees around tuplesort usage.

  2. Allow the built-in ordered-set aggregates to share transition state.

  3. Avoid integer overflow while sifting-up a heap in tuplesort.c.

  4. Allow avoiding tuple copy within tuplesort_gettupleslot().

  5. Remove should_free arguments to tuplesort routines.

  6. Fix use-after-free around DISTINCT transition function calls.

  7. Reuse abbreviated keys in ordered [set] aggregates.