On Thu, May 02, 2013 at 01:01:20PM +0200, Paolo Bonzini wrote: > Il 02/05/2013 12:35, Michael S. Tsirkin ha scritto: > > > I was going to keep these for 1.6, but since they have been tested > > > by the user and myself, and they fix a regression, it can be worthwhile > > > to include these in 1.5. > > > > I remember there were some tricky issues last time > > we tried something like this for PCI. > > PCI is already post-order though, isn't it?
For example, it resets the irq count before. I really don't know why, and the reverse looks more correct but let's tread carefully for 1.5. For 1.6 maybe we'll be able to finally get rid of the irq count. > > Let's find an LSI specific solution for 1.5 please. > > Ok. > > Paolo