On Sun, May 20, 2012 at 12:39:13PM -0400, Kevin O'Connor wrote: > On Sun, May 20, 2012 at 07:25:40PM +0300, Avi Kivity wrote: > > On 05/20/2012 07:16 PM, Kevin O'Connor wrote: > > > > Here we in agreement, and I was against patching till it was > > > > unavoidable, > > > > but than pci hotplug started using it, and afterwards processor > > > > definitions, so no point in avoiding it now by using inferior methods. > > > > > > I agree as well. > > > > > > What's the background to needing to have dynamic S3/S4 definitions? > > > (Why will some qemu instances be able to sleep and not others?) > > > > > > > Backwards compatibility. qemu has a -M machine-type option that expose > > an old qemu's guest-visible attributes. If an old qemu didn't support > > S3, then -M old shouldn't either. > > The DSDT has claimed S3, S4, and S5 support since SeaBIOS has > supported ACPI. What's the background to the requirement to stop > claiming support for it? > Not all guests have working S3/S4 implementation. We want management to be able to disable S3/S4 for such guests. S4 value changes since we want to distinguish between S4 and S5 in qemu.
-- Gleb.