On 29.03.2018 08:33, Janosch Frank wrote: > On 28.03.2018 22:05, Radim Krčmář wrote: >> 2018-03-28 09:39+0200, Thomas Huth: >>> Hi Radim & Paolo, >>> >>> the following changes since commit ebc1b903f12dab0921c055712131a54b7e37eea1: >>> >>> x86: fix compiler and assembly warnings (2018-03-23 18:58:41 +0100) >>> >>> are available in the git repository at: >>> >>> https://github.com/huth/kvm-unit-tests.git tags/s390x-2018-03-28 >>> >>> for you to fetch changes up to 02ce12c322b7ebc8b9d46292d28678e444d4336c: >>> >>> s390x: Add simple guarded storage tests (2018-03-28 09:08:39 +0200) >>> >>> ---------------------------------------------------------------- >>> New unit tests for s390x >>> ---------------------------------------------------------------- >> >> Pulled, thanks. >> >>> Janosch Frank (8): >>> s390x: Add sthyi tests >>> s390x: Add privileged operation exceptions handling >> >> Btw. this test fails on tcg -- is my qemu too old? >> (FAIL: skey: chg bit test) > > tcg uses the qemu skey device, right? > > That device has absolute no idea about what memory was changed > in-between setting and retrieval of the key. Hence it only stores the > key that has been given to it and doesn't set change and reference bits. > > Adding support would make it necessary to set the change bit for an skey > with every cpu store to the associated page. This might be quite slow... > Guess it is sufficient to set the storage key bit when we have have a MMU write fault. Changing the storage key will require to flush the MMU entry. But that's far from trivial. s390x TCG has basically no storage key support yet (e.g. access rights not checked ...). So we will have to live with the failing test for some time - or alternatively only execute it for KVM (which is somewhat cheating). -- Thanks, David / dhildenb
