On Mon, 13 Jul 2015 18:06:45 +0200 Andreas Färber <afaer...@suse.de> wrote:
> Found it. So the problem *is* different from what I understood! It's not > directly attached to /machine by s390x code, but rather instantiated via > -device by the user. > > Peter C. suggested you to do it in realize, which affects all devices. > > The solution would be to instead either do the reset registration in > qdev-monitor.c, where it's specific to devices that do not have a bus > and on /machine/peripheral or /machine/periph-anon are not managed by a > parent, or to add a further check here in realized. Right now I can only > think of a hot-plug flag...? Not sure about unrealizing in the > qdev-monitor case, but I think we can ignore that in this case? This sounds not like something we'll want to do during hardfreeze, though. I'll just fall back to the original patch that registered a reset so this works for now, and we can think of a more generic solution later.
