On Wed, Dec 07, 2022 at 11:44:50PM +0000, Sean Christopherson wrote:
> On Wed, Dec 07, 2022, Oliver Upton wrote:
> > MEM_REGION_TEST_DATA is meant to hold data explicitly used by a
> > selftest, not implicit allocations due to the selftests infrastructure.
> > Allocate the ucall pool from MEM_REGION_DATA much like the rest of the
> > selftests library allocations.
> >
> > Fixes: 426729b2cf2e ("KVM: selftests: Add ucall pool based implementation")
>
> Not that it really matters because no one will backport this verbatim, but this
> is the wrong commit to blame. As of commit 426729b2cf2e, MEM_REGION_DATA does not
> exist. And similarly, the common ucall code didn't exist when Ricardo's series
> introduced MEM_REGION_DATA.
>
> $ git show 426729b2cf2e:tools/testing/selftests/kvm/include/kvm_util_base.h | grep MEM_REGION_DATA
> $ git show 290c5b54012b7:tools/testing/selftests/kvm/lib/ucall_common.c
> fatal: path 'tools/testing/selftests/kvm/lib/ucall_common.c' exists on disk, but not in '290c5b54012b7'
>
> The commit where the two collided is:
>
> Fixes: cc7544101eec ("Merge tag 'kvmarm-6.2' of https://git.kernel.org/pub/scm/linux/kernel/git/kvmarm/kvmarm into HEAD")
Yeah, this Fixes is garbage, apologies.
I imagine Paolo is going to squash some things into the kvmarm merge, so
the Fixes tag ought to be dropped altogether.
> > Signed-off-by: Oliver Upton <oliver.upton@xxxxxxxxx>
> > ---
>
> Fixes nit aside,
>
> Reviewed-by: Sean Christopherson <seanjc@xxxxxxxxxx>
Thanks!
--
Best,
Oliver
