On January 4, 2018, multiple vulnerabilities in the design of modern CPUs were disclosed. Taking advantage of certain processor performance optimizations, these vulnerabilities—named Meltdown and Spectre—make it possible for attackers to coerce applications into revealing the contents of system and application memory when manipulated correctly. These attacks work because the normal privileges checking behavior within the processor is subverted through the interaction of features like speculative execution, branch prediction, out-of-order execution, and caching.
Meltdown was disclosed in CVE-2017-5754. Spectre was disclosed in CVE-2017-5753 and CVE-2017-5715.
For more detailed information, check out the how does meltdown work? and how does spectre work? sections below.
Meltdown and Spectre affect the majority of modern processors. The processor optimizations that are used in these vulnerabilities are a core design feature of most CPUs, meaning that most systems are vulnerable until specifically patched. This includes desktop computers, servers, and compute instances operating in Cloud environments.
Patches to protect against Meltdown are being released from operating system vendors. While updates are also being released for Spectre, it represents an entire class of vulnerabilities, so it will likely require more extensive ongoing remediation.
In cloud and virtualized environments, providers will need to update the underlying infrastructure to protect their guests. Users will need to update their servers to mitigate the impact within guest operating systems.
Full protection against this class of vulnerability will likely require changes in CPU design. In the interim, software updates can provide mitigation against exploits by disabling or working around some of the optimized behavior that leads to these vulnerabilities.
Unfortunately, because these patches affect the optimization routines within the processor, mitigation patches may decrease the performance of your server. The extent of the slowdown is highly dependent on the type of work being performed, with I/O intensive processes experiencing the largest impact.
At the time of writing (January 9, 2018), Linux distributions have started to distribute patches, but no distributions are yet fully patched.
Distributions that have released kernel updates with partial mitigation (patched for Meltdown AND variant 1 of Spectre) include:
Distributions that have released kernel updates with partial mitigation (patched for Meltdown) include:
If your kernel is updated to at least the version corresponding to the above, some updates have been applied.
Operating systems that have not yet released kernels with mitigation include:
Ubuntu 17.04, which is reaching end of life on January 13, 2018 will not receive patches. Users are strongly encouraged to update or migrate.
Warning: We strongly recommend that you update or migrate off of any release that has reached end of life. These releases do not receive critical security updates for vulnerabilities like Meltdown and Spectre, which can put your systems and users at risk.
Because of the severity of this vulnerability, we recommend applying updates as they become available instead of waiting for a full patch set. This may require you to upgrade the kernel and reboot more than once in the coming days and weeks.
To update your servers, you need to update your system software once patches are available for your distribution. You can update by running your regular package manager to download the latest kernel version and then rebooting your server to switch over to the patched code.
Note: This article was written to be generally applicable and platform agnostic. If you are using DigitalOcean as your hosting provider and are running an older Droplet, you may have to perform an extra step before getting started.
DigitalOcean’s legacy kernel management system used externally managed kernels that could be changed in the control panel. If your Droplet uses this system, you will need to configure it to use internal kernel management before continuing (newer Droplets use this system automatically). To check whether you need to update to internal kernels and to learn how to make the switch, read our How To Update a DigitalOcean Server’s Kernel article.
For Ubuntu and Debian servers, you can update your system software by refreshing your local package index and then upgrading your system software:
- sudo apt-get update
- sudo apt-get dist-upgrade
For CentOS servers, you can download and install updated software by typing:
- sudo yum update
For Fedora servers, use the dnf
tool instead:
- sudo dnf update
Regardless of the operating system, once the updates are applied, reboot your server to switch to the new kernel:
- sudo reboot
Once the server is back online, log in and check the active kernel against the list above to ensure that your kernel has been upgraded. Check for new updates frequently to ensure that you receive further patches as they become available.
The Meltdown and Spectre family of vulnerabilities exploit performance-enhancing features within modern processors. A combination of processor features like speculative execution, privilege checking, out-of-order execution, and CPU caching allows read access to memory locations that should be out-of-bounds. The result is that unprivileged programs can be coerced into revealing sensitive data from their memory or accessing privileged memory from the kernel or other applications.
The Meltdown vulnerability works by tricking a processor into reading an out-of-bounds memory location by taking advantage of flaws in a CPU optimization called speculative execution. The general idea works like this:
Meltdown represents a specific vulnerability that can be patched against.
Spectre also works by tricking a processor to misuse speculative execution to read restricted values. The disclosure notices describe two variants with different levels of complexity and impact.
For variant 1 of Spectre, the processor is tricked into speculatively executing a read before a bounds check is enforced. First, the attacker encourages the processor to speculatively reach for a memory location beyond its valid boundaries. Then, like Meltdown, an additional instruction conditionally loads a legal address into cache based on the out-of-bounds value. Timing how long it takes to retrieve the legal address afterwards reveals whether it was loaded into cache. This, in turn, can reveal the value of the out-of-bounds memory location.
Variant 2 of Spectre is the most complicated both to exploit and mitigate against. Processors often speculatively execute instructions even when they encounter a conditional statement that cannot be evaluated yet. They do this by guessing the most likely result of the conditional using a mechanism called branch prediction.
Branch prediction uses the history of previous runs through a code path to pick a path to speculatively execute. This can be used by attackers to prime a processor to make an incorrect speculative decision. Because the branch selection history does not store absolute references to the decision, a processor can be fooled into choosing a branch in one part of the code even when trained in another. This can be exploited to reveal memory values outside of the acceptable range.
Spectre and Meltdown represent serious security vulnerabilities; the full potential of their possible impact is still developing.
To protect yourself, be vigilant in updating your operating system software as patches are released by vendors and continue to monitor communications related to the Meltdown and Spectre vulnerabilities.
Thanks for learning with the DigitalOcean Community. Check out our offerings for compute, storage, networking, and managed databases.
This textbox defaults to using Markdown to format your answer.
You can type !ref in this text area to quickly search our full set of tutorials, documentation & marketplace offerings and insert the link!
Hi all, for all Ubuntu 16.04 users please be aware that the kernel 4.4.0-108 is buggy!! https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1742323
Be sure to upgrade to kernel kernel 4.4.0-109 which was also released.
@jellingwood maybe you should update that in the post.
Even after doing this on a CentOS 7.4 Droplet, spectre-meltdown-checker reports:
How do we fix this?
This comment has been deleted
If your droplet won’t switch to the latest kernel after the described updates and your droplet uses “external kernel management” (ie. changing the kernel via the control panel), you might want to consider changing to “Internal Kernel Management” following https://www.digitalocean.com/community/tutorials/how-to-update-a-digitalocean-server-s-kernel
After that you’ll be able to explicitly change to the kernel you want to use.
Ubuntu 16.04 kernel is updated again, or at least before a link to this post was emailed to me today, I had just completed an upgrade to 4.4.0-112
i have no clue what to do… for someone who is not IT at all - this is mind blowing.
Yum update seems to “remove” the old kernel and installs the new one on Centos 7
yet a uname -r shows the old one still.
a yum update shows no new packages waiting.
No idea what went wrong here.
Just to be clear. Is simply changing the Kernel from the admin console good enough? Or do we still have to update the droplet after changing the kernel? Also some of my droplets have internally managed Kernels. What does that mean exactly?
$ sudo apt-get dist-upgrade >>> is failing on Ubuntu 16.04, I get this error:
Err:1 http://download.owncloud.org/download/repositories/stable/Ubuntu_16.04 owncloud-files 10.0.5-1.1 404 Not Found [IP: 138.201.139.178 80]
let us know when we can try again, thanks!
I’m having a bit of a tough time upgrading my kernels. Running CentOS 6, these are pretty old droplets so had to be changed to the GrubLoader kernel in the control panel. After doing that, they boot to a newer kernel, which is great. But it’s not the newest kernel.
If I try running
yum upgrade kernel
it seems to work, but throws outgrubby fatal error: unable to find a suitable template
and doesn’t have a new kernel on reboot. The grub conf file doesn’t even reference the kernel that I am running, so it’s clearly being ignored. Any suggestions? Any hints on how CentOS is actually deciding what kernel to run?