On Mon, Sep 24, 2012 at 06:53:10PM -0400, Cyril Chemparathy wrote:
The existing fixup code runs too early - before we get a chance to
switch over from the boot-time sub-4G alias to the real physical address
space. Readability aside, part of the reasoning behind the C
re-implementation was to defer the patch application until a later point
in the boot process.
That said, we could patch once at boot, and then re-patch at switch
over, I guess. Is that what you're recommending?
That would solve the problem, and probably in the least disruptive
manner too, since the existing solution will continue to work, and
we'd only need to fixup the movw instructions if we're building in
support for the rare 'high-lowmem' situation.
We don't even need any different module marking - the 'high-lowmem'
modules don't need their movw instructions fixed.