[Xen-ia64-devel] revisit _OSI optimizations?

[Alex Williamson <alex.williamson@xxxxxx>]

   I'm becoming more concerned about our use of _OSI to detect the guest
OS type for optimizations.  I know this is an idea that I originally
proposed, but it seems to be falling apart on newer upstream Linux
kernels.  To get an upstream kernel working correctly on Xen/ia64 HVM,
it has to be booted with acpi_osi=Linux.  I'm not convinced the patches
I sent out to make this happen automatically will make it upstream and
even if they do, we have a gap in Linux doing the right thing
automatically.

   What are our alternatives?  I'm afraid any attempt to detect the
guest OS type will lead to similar problems as _OSI in the future.  As
this is a performance optimization, I'm leaning towards the need for a
config option in the domain spec file.  If a user wants to tune for
performance, they can add a config option, but the default must work
correctly, if not 100% optimal for all guests.  Are there other ideas or
objections?  Anyone want to volunteer to take on this task?  Thanks,

        Alex

-- 
Alex Williamson                             HP Open Source & Linux Org.


_______________________________________________
Xen-ia64-devel mailing list
Xen-ia64-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-ia64-devel