Hi Ackerley, On Mon, 5 Aug 2024 at 18:19, Ackerley Tng <ackerleytng@xxxxxxxxxx> wrote: > > Fuad Tabba <tabba@xxxxxxxxxx> writes: > > > Add the KVM capability KVM_CAP_GUEST_MEMFD_MAPPABLE, which is > > true if mapping guest memory is supported by the host. > > > > Signed-off-by: Fuad Tabba <tabba@xxxxxxxxxx> > > --- > > include/uapi/linux/kvm.h | 3 ++- > > virt/kvm/kvm_main.c | 4 ++++ > > 2 files changed, 6 insertions(+), 1 deletion(-) > > > > <snip> > > Why do we need a cap for `KVM_CAP_GUEST_MEMFD_MAPPABLE` instead of just > making guest_memfd mmap-able? > > Is this to prevent breaking userspace, because a user might be relying > on guest_memfd being not mmap-able? To be able to check that the ability is there, since it is a new capability not available in Linux 6.9 not 6.10 (i.e., after guest_memfd()) was introduced. Cheers, /fuad
