diff options
author | Toshi MARUYAMA <marutosijp2@yahoo.co.jp> | 2017-12-07 12:22:22 +0000 |
---|---|---|
committer | Toshi MARUYAMA <marutosijp2@yahoo.co.jp> | 2017-12-07 12:22:22 +0000 |
commit | 3481b49a1aea34d32dd9ae9a9bac2f2af4523cc0 (patch) | |
tree | e8d94e71fffa9d95ef679edd44ff43fc673a138c /plugins | |
parent | d9d803abb5424bd12ce83b1da0b05b3fe1577cfe (diff) | |
download | redmine-3481b49a1aea34d32dd9ae9a9bac2f2af4523cc0.tar.gz redmine-3481b49a1aea34d32dd9ae9a9bac2f2af4523cc0.zip |
Merged r17060 from trunk to 3.2-stable (#27516)
mercurial: reject malicious command argument
We've got a security report from the Phabricator team, which basically says
--config and --debugger arguments can be injected anywhere to lead to an
arbitrary command execution.
https://secure.phabricator.com/rPa7921a4448093d00defa8bd18f35b8c8f8bf3314
This is a fundamental issue of the argument parsing rules in Mercurial, which
allows extensions to populate their parsing rules and such extensions can be
loaded by "--config extensions.<name>=". There's a chicken and egg problem.
We're working on hardening the parsing rules, but which won't come in by
default as it would be a behavior change.
This patch adds a verification to reject malicious command arguments as a
last ditch. The subsequent patches will fix the problem in more appropriate
way.
Contributed by Yuya Nishihara.
git-svn-id: http://svn.redmine.org/redmine/branches/3.2-stable@17074 e93f8b46-1217-0410-a6f0-8f06a7374b81
Diffstat (limited to 'plugins')
0 files changed, 0 insertions, 0 deletions