[pgpool-general: 1140] Re: read_startup_packet: incorrect packet length

[Tatsuo Ishii]

> I'm currently running pgpool-3.2.1 on Linux-x86_64.  Back in August I
> reported a bug related to packet lengths:
> http://www.sraoss.jp/pipermail/pgpool-general/2012-August/000896.html
> 
> which was later reported as fixed in 3.2.1.  Earlier today, I'm seeing
> the following in my pgpool log:
> 2012-10-24 11:25:27 ERROR: pid 13179: read_startup_packet: incorrect
> packet length (455421229)
> 2012-10-24 11:25:34 ERROR: pid 31594: read_startup_packet: incorrect
> packet length (455421229)
> 2012-10-24 11:25:42 ERROR: pid 30797: read_startup_packet: incorrect
> packet length (455421229)
> 2012-10-24 11:25:49 ERROR: pid 22163: read_startup_packet: incorrect
> packet length (455421229)
> 2012-10-24 11:25:54 ERROR: pid 22165: read_startup_packet: incorrect
> packet length (455421229)
> 2012-10-24 11:25:54 ERROR: pid 17441: read_startup_packet: incorrect
> packet length (1195725915)
> 2012-10-24 11:26:05 ERROR: pid 19463: read_startup_packet: incorrect
> packet length (-2139750145)
> 2012-10-24 11:26:10 ERROR: pid 19934: read_startup_packet: incorrect
> packet length (369295612)
> 
> Is the bug fix intended to eliminate only the 'out of memory'
> condition, or should the errors be eliminated as well?

The fix only intended to eliminate the 'out of memory'
condition. Because the packet length data is sent from clients and
pgpool has no control on it, the only thing pgpool could do is,
checking the data and refuses to proceed if it is not correct.  Such a
client could be one of followings:

1) attacker

2) pool client because of faulty network hardware

3) buggy client
--
Tatsuo Ishii
SRA OSS, Inc. Japan
English: http://www.sraoss.co.jp/index_en.php
Japanese: http://www.sraoss.co.jp