A while ago I rebooted a Proxmox server in cluster and got this scary message:
DMAR-IR: [Firmware Bug]: ioapic 3 has no mapping iommu, interrupt remapping will be disabled
The hardware is a Supermicro X10DRL-i and the BIOS version is 3.2. The latest BIOS version for that board is version 3.4.
After the message came up, the server didn’t start.
iommu is well documented on Proxmox website, but in spite of reading all the documentation I still don’t completely understand the context, and if I should worry about it. It’s in the section about PCI passthrough. All the information is about the happy path, not what to do when things bomb out. Here is the link:
https://pve.proxmox.com/wiki/PCI_Passthrough#Verifying_IOMMU_parameters
Rather I’ll tell you how I got out of this mess.
First I left the server. I though perhaps it’s busy doing something, so whilst I researched the problem I just let it lie. A lot of panic set in because numerous VMs were offline. After about 20 minutes I gave up. It seemed I had to add some obscure settings but I was scared to break things further. This came up a few times:
GRUB_CMDLINE_LINUX_DEFAULT="quiet intel_iommu=on,relax_rmrr iommu=pt vfio_iommu_type1.allow_unsafe_interrupts=1"
Commands such as this were also found to see the status:
dmesg | grep 'remapping' dmesg | grep -e DMAR -e IOMMU
In the end and after 20 minutes I just rebooted again. Low and behold, the server started!
Let us know in the comments if you ever experience this.
- https://kb.protectli.com/kb/pci-passthrough-vt-d-proxmox-ve/
- https://forum.proxmox.com/threads/change-node-in-clusters-ip-address-pve-7-4.145954/#post-659165
- https://forum.proxmox.com/threads/where-to-add-intel_iommu-on-and-iommu-pt.131592/
- https://forum.proxmox.com/threads/where-to-add-intel_iommu-on-and-iommu-pt.131592/#post-578296