On Mon, Oct 12, 2020 at 11:23 AM Nadav Amit <nadav.amit@xxxxxxxxx> wrote: > > > On Oct 12, 2020, at 11:16 AM, Jim Mattson <jmattson@xxxxxxxxxx> wrote: > > > > On Sat, Oct 10, 2020 at 2:52 AM Nadav Amit <nadav.amit@xxxxxxxxx> wrote: > > > >> I guess that the test makes an assumption that there are no addresses > >> greater than 4GB. When I reduce the size of the memory, the test passes. > > > > Yes; the identity-mapped page used for real-address mode has to be > > less than 4Gb. > > > > I think this can be fixed with an entry in unittests.cfg that specifies -m 2048. > > I prefer to skip the test if the conditions do not allow to run it, as I do > not use unittests.cfg. > > I will send a patch later. It doesn't have to be skipped. We just need a variant of alloc_page() that will dole out a page with a 32-bit address.
