|
From: | Anthony Liguori |
Subject: | Re: [Qemu-devel] [PATCH v2] qom: Introduce object_realize_nofail() |
Date: | Thu, 12 Apr 2012 16:57:59 -0500 |
User-agent: | Mozilla/5.0 (X11; Linux x86_64; rv:11.0) Gecko/20120310 Thunderbird/11.0 |
On 04/12/2012 04:50 PM, Peter Maydell wrote:
On 12 April 2012 22:24, Anthony Liguori<address@hidden> wrote:The end goal shouldn't be s/qdev/object/g. That doesn't make things better by just using a new shiny infrastructure. Splitting device initialization into two stages is the fundamental reason for introducing QOM in the first place.It might be yours, doesn't mean it's everybody else's :-)
But clearly my own selfish goals are the only thing that matters :-)
My main desire from shifting to QOM is named connections between devices and killing off the strict hierarchy of qbus buses;
We're pretty much there, no? I think the only thing really left is introducing Pins and qomifying MemoryRegions.
I haven't been particularly bitten by device init issues.
You will be... In order to let a user make those connections you'll need a mechanism to set link properties before realize but after instance_init.
Regards, Anthony Liguori
-- PMM
[Prev in Thread] | Current Thread | [Next in Thread] |