QEMU exited with code 1

[LokiPass]

Hi,
I have installed proxmox and everything was working fine until i tried to passthrough LSI Card 9211-8i IT for SAS drives. When I create VM and start it without LSI passthrough, it starts as expected. When I passthrough LSI Card and start VM, it fails with error “start failed, QEMU exited with code 1”..
Do you have any advice?

 

[Dunuin]

Did you checked the PCI passthrough article in the wiki? Its mostly about GPUs but there are gernal hints too.

 

[leesteken]

Look at the task log and syslog or journalctl to see if there is more information about the actual problem.

 

[LokiPass]

Yes I've added GRUB_CMDLINE_LINUX_DEFAULT="quiet intel_iommu=on" to /etc/default/grub and vfio, vfio_iommu_type1, vfio_pci, vfio_virqfd to /etc/modules.
Thanks, I will check journalctl and syslog for more info.

 

[LokiPass]

Thanks guys I should have remembered to look up logs, it was failing with the error "out of memory". I overcommitted RAM to this VM.
It's interesting that when I wasn't adding PCI passthrough it wasn't failing.

 

[Dunuin]

As soon as you use PCI passthrough the RAM is fixed and only that VM can use the RAM. Try to avoid RAM overprovisiong, that isn't working well in general...

 

[leesteken]

When using PCI passthrough, those devices can do Direct Memory Access (DMA) and all VM memory needs to be locked into actual host RAM (pinned memory). This is the same reason that memory ballooning does not work for VMs with passthrough. The device can read but also write to the VM's memory at any time, therefore it cannot be returned to the host or used by other VMs. This is the price we pay for direct access to PCI(e) devices.

 

[mzahor]

Thanks guys I should have remembered to look up logs, it was failing with the error "out of memory". I overcommitted RAM to this VM.
It's interesting that when I wasn't adding PCI passthrough it wasn't failing.

Man, you saved my day!!! Had exactly the same issue, was pulling my hair off, nothing worked and just getting exit code 1. After limiting the memory size to 12Gb instead of 16 (the whole server memory), it finally worked. Where did you find those logs though? I tried journalctl but didn't see the error message memory.