In D19495#449190, @rgrimes wrote:

In D19495#449173, @darius.mihaim_gmail.com wrote:

I have applied the feedback @jhb gave, except for the #ifdef guards. Will work on it after rebasing with a newer master.

Please delay the #ifdef guards, as per the discussion in the 2019/06/20 bhyve call we may of found an alternative, using a project branch and rebasing the git to that.

In D10213#323447, @mmacy wrote:

@mihai Before this can proceed this review needs to be broken up in to at _least_ 3 pieces

1. MI/MD refactoring of sys/amd64/vmm - you need to identify all the common interfaces that are shared and move them out of amd64, this means you need to define a clean shared API where possible, e.g. starting / stopping guest / setting nvcpus / etc

2. MI/MD refactoring of libvmmapi

We tried to do this before, but we analyzed the interfaces, we saw that they are x86 dependent (memseg set, etc). We would end up with a common interface only for starting/stopping/running.

We don't have any implementation for them in arm and most of them have some x86 specific bits (like registers). This is why we choose to only make what we have machine independent. And when new features appear for ARM we would take care to make them as general as possible (e.g. virtio.c -> which is WIP)

Can you extend what means MD/MI?

Any update on this?

Any update on this?

Any update on this?

Alex Ivan update the diff using Andrew recommendations (parent or maintenance entry)

Updated with more context.

Add HMAIR registers.

• modified cast
• fix tab/spaces

Ordered entries. If they are not order, please provide the methodology of ordering them.

Can we commit this?

I've created three new patches containing:

• sysreg bits
• generic_timer bits
• gic bits.

This new version of the patch fixes comments from Andrew Turner and Peter Grehan:

• create platform dependent code for bhyve (no more bhyvearm/libvmmapiarm duplicate directory)
• create a device for the vgic in order to user the getter methods from the gic device
• added timer virtualization
• completed vGIC virtualization

I've put some questions into the comments. As soon as I get answers to those, I will solve them and UPDATE the commit diff.

I'm ok with commiting it.

The comment "/* Leave HYP mode */" isn't consistent. We are checking if we are in HypMode only.

Uploaded a file instead of copy/paste.

Update the diff acording to the last review.

I've applied all the recomandations from Andrew, besides the "struct arm_boot_params *" one.

Forgot to add the new file sys/arm/arm/hypervisor-stub.S