> Did this work before? Yes 

Probably the easiest is to bisect then?  I don't personally have time to do this though...

I can revert this if it's causing an issue -- but it should only do that if you're using an upstream kernel (or one that doesn't have a chrome os config in the kernel tree) 

any reason you're not using 4.4 like regular kevin?

kevin64 does use the regular 4.4 kernel:

equery-kevin64 l '*kernel*'
 * Searching for *kernel* ...
[I-O] [  ] sys-kernel/chromeos-kernel-4_4-4.4.157-r1682:0

So I think there must be some other unexpected issue here?

Regarding to #3, I've just reverted this CL locally and after that it works again.

re #8 - that is strange, that CL should only affect kernels which don't have a chrome os style config made with the scripts under chromeos/scripts using snippets from chromeos/config/ -- and if you're using 4.4 you should be using the chrome os config that we keep in the kernel.

I'll try to build kevin64 and see if I can figure out what's going on

it does look like kevin64 is using chrome os kernel 4.4 and using the same config as kevin -- the only difference I noticed is that kevin has USE=kvm_host set, which turns on some different kernel configs.  However, none of those should affect the display.

So I'm wondering -- are you building the kernel from your local checkout (are you cros worked-on chromeos-kernel-4_4 ?) 

if you are doing that then maybe using a different kernel or have an upstream or other non-chromeos kernel checked out? 

Thanks for your comment. I've been working on re-enable kvm-host on kevin64. We will see whether it has any effect related to this issue. 

No, we aren't. We use the 'official' kernel 4_4 package for the building.

I've looked at this issue, I checked the kernel serial output, the init process was killed by another reason.
'/sbin/init: error while loading shared libraries: libdbus-1.so.3: cannot open shared object file: No such file or directory'

That's not clear to me why it is gone when I reverted this, because I don't see there is no relationship with this change. So I think this issue could be closed.