Re: [Xen-devel] [PATCH 2/10] PM interface between dom0 and Xen

[Keir Fraser <keir@xxxxxxxxxxxxx>]

Patches 1 and 2 are now in with only very slight changes.

However, I have not yet enabled the new platform hypercalls. I believe they
can be slimmed down significantly.

 1. We don't need set_acpi_sleep() at all. The values it provides are all
quite easily found in the fixed-format FADT and FACS structures. We don't
have the code for finding and parsing those tables in Xen currently, and we
don't want to pull in the whole Linux generic mechanism! *But* I expect it
would not be hard to put together a small routine to discover waking_vector,
waking_vector_length, pm1{a,b}_{evt,vnt}_port, based on the Linux code
(which we want to crib from since it does all the right things for different
ACPI versions) but written from scratch to be tailored for extracting just
those values that we require for S-state support.

 2. Enter_acpi_sleep() does not require the magic video_{flags,mode} fields.
If this is just to support "acpi_sleep=" cmdline option, we can support that
option directly in Xen and cut dom0 out entirely.

I'll carry on looking through the remainder of the patches, since I can
apply those in parallel with any changes being made to this patch 2/10.

 -- Keir

On 27/6/07 14:30, "Tian, Kevin" <kevin.tian@xxxxxxxxx> wrote:

> Define pm interfaces between dom0 and Xen, with one
> to register sleep info and the other for triggering
> sleep state. "acpi_sleep=s3_mode/s3_bios" option is
> also supported by piggybacking video flag/mode to
> xen at trigger point.
> 
> Signed-off-by Ke Yu <ke.yu@xxxxxxxxx>
> Signed-off-by Kevin Tian <kevin.tian@xxxxxxxxx>


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel