[vfio-users] VM Boot Hang After Unclean Shutdown w/ ASMedia ASM1042A USB 3.0 PCI-e

[Adam Zegelin]

Martin,

Thanks for your mail and "me too" on this issue.

This perhaps sounds like a wider issue with ASMedia USB 3.X
controllers not adhering to spec.
Hopefully the issues can either be worked around in software, or
ASMedia is kind enough to offer a firmware update
(assuming that its a firmware level issue).

My questions for the rest of the list/wider community are:

How does an OS usually "de-init" an XHCI controller?
(i.e, What is so special about a graceful vs. ungraceful shutdown that
causes the controller to fail on the next boot?)
Is this de-init manufacturer/driver specific?
Is there a reason (apart from "not implemented yet") that
qemu/kvm/<whatever> couldn't do the same on guest termination?

Since this issue is reproducible -- `killall qemu` + reboot vm -- is
there a way that I can offer my assistance to those
who have more intricate knowledge of how all the moving parts fit
together? I would like to help.

I'm a software developer by day but my understanding of
device/firmware/bootloader/VM level stuff is very hand-wavy/high-level
at best.

- Adam


On Sat, Jan 14, 2017 at 9:17 PM, Martin Schrodt <martin at schrodt.org> wrote:
> Hi Adam,
>
> I just want to add myself to the list. In my case, I have a single
> ASMedia 1142 passed through:
>
> 04:00.0 USB controller: ASMedia Technology Inc. ASM1142 USB 3.1 Host
> Controller
>
> In my case, there's no ACS override, but I still experience the exact
> same problem. As long as the VM is shut down gracefully, I can boot the
> VM again, in case of crash or hard reset it won't boot anymore until
> host restart.
>
> I spent some time ago to somehow remove/reset/whatnot the controller in
> such a case - to no avail.
>
> I remember Alex Williamson being shocked, *shocked* I say, about ASMedia
> devices not behaving. The case was that when my controller was not
> passed through, but used by the host, it would prevent the system to
> properly suspend to RAM.
>
> I'd be very interested in knowing a way to archieve proper reset of the
> controller, but I fear there simply is none.
>
> Martin
>
> _______________________________________________
> vfio-users mailing list
> vfio-users at redhat.com
> https://www.redhat.com/mailman/listinfo/vfio-users