Should you do microcode updates?
A colleague and I were discussing my recent rant re a RedHat-issued kernel update breaking VMs on all my AMD systems. Notwithstanding the fact that my RedHat bug report about the issue is set to “private” and Bugzilla (1990 called wants its UI back, to hang out with Miranda IM) won’t let me change it, he asked me “we’ve applied the latest microcode, right?”
I said, “What microcode?”
So began my schooling in the murky world of CPU microcode - what seems to be “realtime patches” to the hardcoded firmware in your CPU processor. I’ve probably misunderstood much of this, but I imagine something like this:
What are microcode updates?
Let’s say we have a CPU with some buggy code, we’ll call it the AME Venom. The CPU will always calculate 2+2=5. Disaster. Shortly after this is discovered, OS vendors release a microcode update which tells the operating system, “When you boot, if you’re running on an AME Venom, then ignore what the CPU tells you about 2+2. The answer is actually 4”
OS vendors will routinely issue updated microcode packages, which (when installed) will update your initrd, so that on boot, your kernel will load this “software band-aid” and not become confused by faulty old CPU code.
Should you apply microcode updates?
Well, it depends on what you want to protect against. The latest in a series of horrible microcode updates from Intel will either suck your performance, or ransom the security of your CPU, calling it a “feature”. And the dreaded SPECTRE/MELTDOWN vulnerability only impacts you if you’re already allowing untrusted miscreants to run code (or VMs) on your platforms.
But, if like me, you’re running some 3+-year-old AMD platforms which have never had a microcode update, you might find yourself unable to boot newer kernels ?, without applying the dis_ucode_ldr fix as a kernel boot argument.
How do you apply microcode updates?
Under a RedHat-based distribution, just run
yum install microcode_ctl. Under Debian/Ubuntu, run
microcode.ctl. And reboot. Job’s done.
To confirm your version of microcode on your distribution/CPU platform, run
cat /proc/cpuinfo | grep microcode.
You’ve totally misunderstood microcode
Probably. I’ve only known about it for a few hours. Let’s discuss. See below :)