Trustly PostgreSQL Data Corruption Bug Bounty Program

From: Joel Jacobson <joel(at)trustly(dot)com>
To: Pg Hackers <pgsql-hackers(at)postgresql(dot)org>
Subject: Trustly PostgreSQL Data Corruption Bug Bounty Program
Date: 2015-06-24 17:15:19
Message-ID: CAASwCXe0bkvZaVBYfDExFpt6UAPvFqs5pE4bVWzB7r=2CjDe7w@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Dear hackers,

The text below has also been published at:
https://trustly.com/en/postgresql-bug-bounty/

You are most welcome to contact me at joel(at)trustly(dot)com or by replying
to this thread, if you have any suggestions on how to improve this
bug bounty program, or if you have any other feedback in general.

Many thanks.

Joel

--

Most people take for granted that PostgreSQL will never corrupt their data,
because the database has such an impressive track record and few users will
ever experience any data corruption problems, even after years of continuous
heavy load.

But just as we shouldn't take our beloved ones for granted, we shouldn't take
the reliability of PostgreSQL for granted. We want to give a small monetary
incentive to the people who look after our beloved PostgreSQL and make sure
that data corruption problems never ever reach a release candiate.

What's the bounty?
Trustly pays USD 1024 per verified data corruption bug in the HEAD of the master
branch of PostgreSQL.

Trustly's bug bounty program is not associated with the official PostgreSQL
project; it's a completely separate initiative having little to do with the
project, except that the bugs we are interested in finding happen to be in the
code the PostgreSQL project works on.

The bug must be demonstrated against the HEAD of the master branch of the
official PostgreSQL git repository, and must not have been reported or
discovered earlier.

Reproducing the bug must not require any types of hardware failure or
unreasonable actions by the administrator, such as turning off fsync or
full_page_writes. However, you are allowed to crash the server at will.

The submitted bug must be reproducible in a unit test and show any of the
following phenomena:
- database not starting up after a shutdown or a crash
- acknowledged commit not recoverable (data silently disappeared, or errors
while trying to read or modify the data)
- data not written to the database appearing in SELECTs (duplicates of the
same row, deleted rows reappear, or garbage data)

You are allowed to assume that PL/pgSQL is available, and bugs which can be
demonstrated to lead to any of the above phenomena only inside PL/pgSQL
functions are still eligible for the bounty, except for bugs which lead to
errors when trying to read or modify data.

- Why?
With this bug bounty program, our hope is to incentivize more people to work
on new clever ways of testing PostgreSQL or to invent other methods capable
of finding data corruption bugs in PostgreSQL.

The objective is to shorten the number of commits between the commit which
introduced a data corruption bug and the commit which fixed it.

If bugs in the master branch are not detected early enough, there is a risk
they will stick around undetected long enough to be included in a released
version of PostgreSQL.

If instead a new data corruption bug introduced in the master branch is fixed
before ever being included in the next release, then no harm is done from a
user perspective, since users who care much about their data only run released
versions of PostgreSQL in production.

- How?
To report a data corruption bug, please first submit it to the PostgreSQL
project by following the instructions at
http://www.postgresql.org/support/submitbug/.
Once the bug has been verified by the PostgreSQL project, go to
http://www.postgresql.org/list/pgsql-bugs/ to locate your bug report and its
message-id, then e-mail your bug report to bug-bounty(at)trustly(dot)com to collect
your reward.

Bounties are awarded at the discretion of Trustly. We only pay individuals.
Please allow up to two weeks for us to verify a found bug.

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Peter Geoghegan 2015-06-24 17:35:20 Re: UPSERT on partition
Previous Message Fabrízio de Royes Mello 2015-06-24 17:12:18 Re: Add CINE for ALTER TABLE ... ADD COLUMN