On Tue, 15 Feb 2022 07:32:12 +0000, Andrew Jones <drjones@xxxxxxxxxx> wrote: > > On Mon, Feb 14, 2022 at 05:01:40PM +0000, Marc Zyngier wrote: > > Hi all, > > > > On Mon, 14 Feb 2022 16:20:13 +0000, > > Alexandru Elisei <alexandru.elisei@xxxxxxx> wrote: > > > > > > Hi Drew, > > > > > > (CC'ing Marc, he know more about 32 bit guest support than me) > > > > > > On Mon, Feb 14, 2022 at 03:24:44PM +0100, Andrew Jones wrote: > > > > I'm not sure it's worth adding another patch on top for that now, though. > > > > By the lack of new 32-bit arm unit tests getting submitted, I'm not even > > > > sure it's worth maintaining 32-bit arm at all... > > > > > > As far as I know, 32 bit guests are still very much supported and > > > maintained for KVM, so I think it would still be very useful to have the > > > tests. > > > > I can't force people to write additional tests (or even start writing > > the first one), but I'd like to reaffirm that AArch32 support still is > > a first class citizen when it comes to KVM/arm64. > > > > It has been tremendously useful even in the very recent past to debug > > issues that were plaguing bare metal Linux, and i don't plan to get > > rid of it anytime soon (TBH, it is too small to even be noticeable). > > > > OK, let's keep 32-bit arm support in kvm-unit-tests, at least as long as > we can find hardware to test it with (I still have access to a mustang). That HW will be with us for a *very* long time, given how popular things like A53 and A72 have been (and still are). > Does kvmtool support launching AArch32 guests? If so, then I suppose we > should also test kvmtool + 32-bit arm kvm-unit-tests. It does. Passing --aarch32 as a parameter results in AArch32 EL1 being selected. The only thing that doesn't work with kvmtool and this flag is the kvmtool-provided init when booting a 32bit kernel directly (we only carry a 64bit version). This should not affect unit tests, which are standalone. Thanks, M. -- Without deviation from the norm, progress is not possible.