Re: [Qemu-devel] [PATCH 2/3] ioh3420: Provide a unique bus name and an interrupt mapping function

From: Markus Armbruster <armbru@redhat.com>
To: Paolo Bonzini <pbonzini@redhat.com>
Cc: Marcel Apfelbaum <marcel.a@redhat.com>,
	Alexey Kardashevskiy <aik@ozlabs.ru>,
	Juan Quintela <quintela@redhat.com>,
	Knut Omang <knut.omang@oracle.com>,
	qemu-devel@nongnu.org, Gonglei <arei.gonglei@huawei.com>,
	"Michael S. Tsirkin" <mst@redhat.com>,
	Igor Mammedov <imammedo@redhat.com>
Subject: Re: [Qemu-devel] [PATCH 2/3] ioh3420: Provide a unique bus name and an interrupt mapping function
Date: Wed, 20 Aug 2014 16:28:13 +0200	[thread overview]
Message-ID: <87r40bw802.fsf@blackfin.pond.sub.org> (raw)
In-Reply-To: <53F4A03F.9050300@redhat.com> (Paolo Bonzini's message of "Wed, 20 Aug 2014 15:18:55 +0200")

Paolo Bonzini <pbonzini@redhat.com> writes:

> Il 20/08/2014 15:03, Markus Armbruster ha scritto:
>>> >
>>> > This is for the root bus, I think it won't help Knut who's trying to
>>> > add devices behind root ports.
>> Read again, more slowly :)
>> 
>> Yes, I null the name of the root bus.  That makes the qdev machinery
>> derive the very same "pcie.0" name via rule 3 instead of rule 1, with
>> the side effect that future (non-root) PCIE buses get different names.
>> In particular, the next one named via rule 3 will be called "pcie.1"
>> instead of "pcie.0", making it actually accessible.
>
> I agree that this is a big improvement.

It's also an ABI break.  I'm not saying we can't do it, just that we
better consider it carefully.