Xen Project discloses serious vulnerability that impacts virtualized servers

virtualization cloud
Credit: Thinkstock

The Xen Project has revealed the details of a serious vulnerability in the Xen hypervisor that could put the security of many virtualized servers at risk.

Xen is a free, open-source hypervisor used to create and run virtual machines. It is widely used by cloud computing providers and virtual private server hosting companies.

The security vulnerability, which is being tracked as CVE-2014-7188 and was privately disclosed to major cloud providers in advance, forced at least Amazon Web Services and Rackspace to reboot some of their customers' virtualized servers over the past week.

The issue allows a virtual machine created using Xen's hardware-assisted virtualization (HVM) to read data stored by other HVM guests that share the same physical hardware. This breaks an important security barrier in multi-tenant virtual environments.

A malicious HVM guest can also exploit the flaw to crash the host server, the Xen Project said in a security advisory published Wednesday.

The vulnerability only affects Xen running on x86 systems, not ARM, and does not impact servers virtualized with Xen's paravirtualization (PV) mode instead of HVM.

Even so, the issue is likely to affect a very large number of servers. Amazon was forced to reboot up to 10 percent of its Elastic Cloud Compute (EC2) servers over the last several days in order to apply the patch and Rackspace's similar effort affected a quarter of its 200,000 customers.

Amazon scheduled its reboots so they didn't affect two regions or availability zones at the same time.

"The zone by zone reboots were completed as planned and we worked very closely with our customers to ensure that the reboots went smoothly for them," the company said Wednesday in a blog post.

Things did not went as smooth for Rackspace whose CEO, Taylor Rhodes, admitted in an email sent to customers Tuesday that the company "dropped a few balls" in the process of dealing with the vulnerability.

"Some of our reboots, for example, took much longer than they should," Rhodes said. "And some of our notifications were not as clear as they should have been. We are making changes to address those mistakes."

The march toward exascale computers
View Comments
Join the discussion
Be the first to comment on this article. Our Commenting Policies