[pgpool-hackers: 838] Re: [pgpool-committers: 2498] pgpool: Fix for pgpool-II emits a fatal error message when clear text p

[Muhammad Usama]

Hi Tatsuo

Thanks for catching this. It somehow slipped through the cracks while
testing. The segfault was already present there and not caused by last
check-in.
Although I was able to reproduce the crash by following the steps you
mentioned, but I was getting the different stack trace for the crash. I
have found the problem and pushed the fix for that. Now the mentioned
scenario is working fine and I am getting the clean regression.
Can you confirm if it also fixes the segfault at your side.

Thanks
Best regards!
Muhammad Usama


On Thu, Mar 26, 2015 at 4:57 AM, Tatsuo Ishii <ishii at postgresql.org> wrote:

> Hi Usama,
>
> Thank you for the fix.
>
> However the fix has a problem: when if pgpool-II already has a
> connection cache to backend, pgpool-II segfaults. To easily reproduce
> the problem, you set num_init_children to 1 which forces pgpool-II to
> reuse same child process instance for incoming connections.
>
> Here is a pgpool log:
>
> 2015-03-26 08:52:56: pid 14970: LOG:  new connection received
> 2015-03-26 08:52:56: pid 14970: DETAIL:  connecting host=[local]
> 2015-03-26 08:55:18: pid 14957: WARNING:  child process with pid: 14970
> was terminated by segmentation fault
>
> Here is the stack trace:
>
> Program received signal SIGSEGV, Segmentation fault.
> MemoryContextDelete (context=0x25f34c0) at ../../src/utils/mmgr/mcxt.c:184
> 184             MemoryContextDeleteChildren(context);
> (gdb) bt
> #0  MemoryContextDelete (context=0x25f34c0) at
> ../../src/utils/mmgr/mcxt.c:184
> #1  0x000000000046ab25 in MemoryContextDeleteChildren (context=0x25f34c0)
>     at ../../src/utils/mmgr/mcxt.c:213
> #2  0x000000000046aae9 in MemoryContextDelete (context=0x25f34c0) at
> ../../src/utils/mmgr/mcxt.c:184
> #3  0x000000000046ab25 in MemoryContextDeleteChildren (context=0x25f34c0)
>     at ../../src/utils/mmgr/mcxt.c:213
> #4  0x000000000046aae9 in MemoryContextDelete (context=0x25f34c0) at
> ../../src/utils/mmgr/mcxt.c:184
> #5  0x000000000046ab25 in MemoryContextDeleteChildren (context=0x25f34c0)
>     at ../../src/utils/mmgr/mcxt.c:213
> #6  0x000000000046aae9 in MemoryContextDelete (context=0x25f34c0) at
> ../../src/utils/mmgr/mcxt.c:184
> #7  0x000000000046ab25 in MemoryContextDeleteChildren (context=0x25f34c0)
>     at ../../src/utils/mmgr/mcxt.c:213
> (repeats forever)
>
> Best regards,
> --
> Tatsuo Ishii
> SRA OSS, Inc. Japan
> English: http://www.sraoss.co.jp/index_en.php
> Japanese:http://www.sraoss.co.jp
>
> > Fix for pgpool-II emits a fatal error message when clear text password
> auth is used.
> >
> > The problem was introduced in pgpool-II after inclusion of exception
> manager.
> > EOF on the frontend connection during client authentication should not
> be treated
> > as an error when the password authentication method is used. psql
> disconnects
> > from the server in such case to get the password from the terminal,
> > reconnects and sends the read password to server.
> >
> > The check-in also enhances the error messages for authentication failures
> > by concatenating backend errors to pgpool error report.
> >
> > Branch
> > ------
> > master
> >
> > Details
> > -------
> >
> http://git.postgresql.org/gitweb?p=pgpool2.git;a=commitdiff;h=46a3982714b19ceaab1612b00b85ea31a053c25d
> >
> > Modified Files
> > --------------
> > src/auth/pool_auth.c    |   39 +++++++++++++++++++++++++++++++++++++--
> > src/include/pool.h      |    2 ++
> > src/protocol/child.c    |   11 ++++++++++-
> > src/utils/pool_stream.c |    5 ++++-
> > 4 files changed, 53 insertions(+), 4 deletions(-)
> >
> > _______________________________________________
> > pgpool-committers mailing list
> > pgpool-committers at pgpool.net
> > http://www.pgpool.net/mailman/listinfo/pgpool-committers
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.sraoss.jp/pipermail/pgpool-hackers/attachments/20150326/5a24dbf3/attachment.html>