Move the __jiffies and _IdleTimer variables to the .data16 segment.
This ensures that if do_idle is invoked very early that it will not
jump off into a non-existent idle routine.

Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>

Resolved Conflicts:
	com32/modules/Makefile
	core/comboot.inc
	core/cpuinit.inc
	core/idle.inc

Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>

Do the actual idling in protected mode.  This both allows PM code a
more efficient interface, but also handles bugs in HVM implementations
which don't handle HLT in real mode.

Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>

This fixes a small inaccuracy in memdisk:

get_disk_image_geometry() does not use MBR data to detect disk
geometry, since it checks MBR signature at wrong offset.

This normally does not cause problems, because memdisk guesses a
reasonable value, if no MBR is found.

Signed-off-by: Bodo Stroesser <bstroesser at ts.fujitsu.com>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>

Remove do_idle from pollchar; it is highly inappropriate while, for
example, checking for a keypress during disk activity.  Furthermore,
make sure we reset the idle counter on events that indicate actual
disk activity.

Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>

Add a very simple com32 program to dump a single CPUID leaf.

Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>

Xen HVM when used with the "vmxassist" program crashes if HLT is
executed in real mode.  Detect a version of Xen old enough to support
"vmxassist", and if it is found, then disable HLT.

See: https://bugzilla.redhat.com/show_bug.cgi?id=570496

Reported-by: Andrew Jones <drjones@redhat.com>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>

Skip leading garbage in the version and copyright strings reported via
the comboot API.  Clean up fixes for those, and document the situation.

Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>