RHEL6 and libvirtd affecting net.ipv4.ip_forward
Shawn Wells
shawn at redhat.com
Thu Dec 12 03:31:41 UTC 2013
On 12/11/13, 12:10 PM, Steve Grubb wrote:
> On Wednesday, December 11, 2013 04:55:39 PM Robert Sanders wrote:
>> >Morning again,
>> > While working a bit more on the sysctl issue with net.ipv4.ip_forward and
>> >net.ipv4.conf.all.accept_redirects found another interesting tidbit. My
>> >boxes are stock installs, which includes the libvirtd service as an enabled
>> >service. I noticed that my /etc/sysctl.conf file explicitly had
>> >'net.ipv4.ip_forward = 0', but the output of 'sysctl net.ipv4.ip_forward'
>> >shows a running value of '1'. This makes sense, as libvirtd I believe
>> >needs to be able to forward packets potentially between any virtual NICs on
>> >the system. My question - is there a STIG that requires this service to be
>> >disabled or not installed. If not, then RHEL-06-000082 will never be
>> >satisfied. Bear in mind, I'm still working from the published RHEL6 STIG,
>> >not the SSG document at this time.
> In my opinion, there should be a STIG specifically for virtualization. For
> example, you absolutely_must_ have the clean traffic ebtables rules loaded or
> you have potential for all kinds of mischief.
>
> -Steve
>
>
> https://www.berrange.com/posts/2011/10/03/guest-mac-spoofing-denial-of-service-and-preventing-it-with-libvirt-and-kvm/
RHEL6 STIG is meant as a Server, not hypervisor. Feedback as the RHEL6
content is applied to a hypervisor would be most welcome; at some point
we'll make a stig-rhel6-kvm.
IIRC, DISA FSO was socializing a "Virtualizer SRG" a few months back. I
can't seem to find a copy via google, but recall conversations of having
RHEL6+KVM, RHEV-H, and OpenStack Nova follow those requirements.
I'm not sure how VMWare handled their ESXi STIG; I think they went
through the OS+Network+AppServer SRGs and cherry picked things. Red Hat
may do the same for our virt technologies until the Virtualizer SRG is out.
More information about the scap-security-guide
mailing list