Implement MsrList and Msrs as FamStructWrappers

[acatangiu]

Description of changes

kvm_msr_list and kvm_msrs each contain a flexible array member and were forcing users of kvm-ioctls crate to work with unsafe code.
Use the kvm-bindings CpuId, MsrList and Msrs safe wrappers over them and change our ioctl functions to use these safe wrappers.

[jiangliu]

Could we implement this in kvm-bindings crate? It should be simpler and natural to implement mechanism to access syscall related structs within the kvm-bindings crate.

[andreeaflorescu]

@jiangliu I don't think that implementing these traits in the kvm-bindings crate is a good idea at the moment. There are a few reasons for that:

1. kvm-bindings still separates clearly the autogenerated code from the one that is not autogenerated. I would prefer to keep it this way for now.
2. we are in discussions about merging/collaborating with ketuvim (Collaboration / Merge with Ketuvim community#62). Probably it will become more clear after the rust-vmm meetup what is the plan there.

While I believe that it's not the best thing now to move this functionality in kvm-bindings, we can discuss about it again once the rust-vmm components and their separation are more obvious.

What do you say?

[jiangliu]

@jiangliu I don't think that implementing these traits in the kvm-bindings crate is a good idea at the moment. There are a few reasons for that:

1. kvm-bindings still separates clearly the autogenerated code from the one that is not autogenerated. I would prefer to keep it this way for now.
   When enabling serde for kvm-bindings, I ran into the same issue to add non-auto-generated code to the kvm-bindings crate. In some sense, we are enhancing the bindgen tool to better support FAM structs, so
   propose to implement in the kvm-binding crate.
   It's great to collaborate with the ketuvim project, which may help to solve all these issues.

2. we are in discussions about merging/collaborating with ketuvim (rust-vmm/community#62). Probably it will become more clear after the rust-vmm meetup what is the plan there.

While I believe that it's not the best thing now to move this functionality in kvm-bindings, we can discuss about it again once the rust-vmm components and their separation are more obvious.

What do you say?

[acatangiu]

After some thought, I agree that the FamWrappers over the kvm FAM structures belong in the same crate as the structs themselves (kvm-bindings). Doing that also simplifies the implementation since we can directly implement the external FamStructWrapper trait on a local struct.

I've opened rust-vmm/kvm-bindings#8 that exemplifies how that'd look.
The PR adds a new module in that crate that doesn't touch the generated files, it just imports the structs from there, so that's nice. The downside is that kvm-bindings would depend on vmm-sys-utils.

[acatangiu]

The PR is updated to use the kvm-bindings safe wrappers. Please review when you get the chance.