Packer: qemu builder net_device validation too aggressive

Created on 24 Mar 2019 · 6Comments · Source: hashicorp/packer

I think the hardcoded list of allowed values for net_device in qemu builds may have lagged behind the available virtual NIC's that qemu now offers. For example, virtio-net-device is offered by qemu-system-aarch64, but this value is rejected by packer for net_device.

bug buildeqemu community-supported plugin good first issue

Source

mcandre

Most helpful comment

I think the chain of commands that can be used to get (more or less definite?) list is something like this

// get all machine types
qemu-system-* -machine help

// get all accelerators
qemu-system-* -accel help



md5-eb94101bcd858dbce629be740afba1db



qemu-system-* -machine type={machine-type-from-above},accel={accel-from-above} -device help



md5-492def491e767bab526c7ae16b80ebbc



$ qemu-system-arm -machine type=xyz
qemu-system-arm: -machine type=xyz: unsupported machine type
Use -machine help to list supported machines

So as long as Packer can transparently pass that error to the user I think we could just avoid the static validation altogether? 🤔

This is on QEMU 3.1.0 - it may be worth checking 2.x.x too, but I doubt it will be much different.

radeksimko on 22 Apr 2019

👍4

All 6 comments

Hey, thanks. Do you know of a list of allowed values I could refer to in order to update the validation?

SwampDragons on 25 Mar 2019

I think the chain of commands that can be used to get (more or less definite?) list is something like this

// get all machine types
qemu-system-* -machine help

// get all accelerators
qemu-system-* -accel help



md5-eb94101bcd858dbce629be740afba1db



qemu-system-* -machine type={machine-type-from-above},accel={accel-from-above} -device help



md5-492def491e767bab526c7ae16b80ebbc



$ qemu-system-arm -machine type=xyz
qemu-system-arm: -machine type=xyz: unsupported machine type
Use -machine help to list supported machines

So as long as Packer can transparently pass that error to the user I think we could just avoid the static validation altogether? 🤔

This is on QEMU 3.1.0 - it may be worth checking 2.x.x too, but I doubt it will be much different.

radeksimko on 22 Apr 2019

👍4

For now, it may better to turn off validation and allow good builds to succeed, with proper validation reintroduced later. That would represent a more practical false positive / false negative situation. The current code makes it hard to manage qemu guests.

mcandre on 26 Jun 2019

That's fair. I'm happy to stop validating as long as the error bubbles up properly.

SwampDragons on 26 Jun 2019

I ran into this issue myself with a powerpc machine since the sungem adapter isn't available in the default list.

soul4soul on 8 Nov 2019

I'm going to lock this issue because it has been closed for _30 days_ ⏳. This helps our maintainers find and focus on the active issues.

If you have found a problem that seems similar to this, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.