No core dump on dbus-daemon crash

Simon McVittie simon.mcvittie at collabora.co.uk
Mon Mar 16 13:31:23 PDT 2015


On 16/03/15 16:43, Jean-Pierre.Bogler at continental-corporation.com wrote:
> I'm currently hunting a bug where the dbus-daemon (version 1.6.4)
> sporadically is killed
> by a SIGSEGV during system start.

dbus 1.6.4 is about 2.5 years old and is vulnerable to CVE-2012-3524,
CVE-2013-2168, CVE-2014-3477, CVE-2014-3533, CVE-2014-3532,
CVE-2014-3639, CVE-2014-3638, CVE-2014-3637, CVE-2014-3636,
CVE-2014-3635, CVE-2014-7824, CVE-2015-0245. It also does not mitigate
third-party flaws like CVE-2014-8148, and has an assortment of known
bugs that are not security vulnerabilities which I have not listed here.

Please consider not doing that. If you got dbus 1.6.4 from a specific
vendor, either upgrade or get support from them.

> Unfortunately, I don't get a core dump
> when the daemon
> is started by the user "messagebus".

This is probably because, in this configuration, it has called setuid(),
triggering code paths in the kernel intended to protect processes that
are privilege boundaries.

> If I remove "<user>messagebus</user>" from the system(1).conf. The crash
> dumps are written!

This also means your dbus-daemon runs as root, which is a security and
sanity risk. It is not really designed to run like that; you have been
warned.

I suggest looking into systems that use the kernel's support for writing
core dumps to a pipe, such as corekeeper.

    S



More information about the dbus mailing list