The calculation of vmin in overflow_general() was bogus, causing silliness like ~80h being warned about in a byte context. Reported-by: C. Masloch <pushbx@38.de> Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>