On Fri, 2 Dec 2022 12:20:34 +0100 Janosch Frank <frankja@xxxxxxxxxxxxx> wrote: > On 12/2/22 11:39, Nico Boehr wrote: > > Quoting Janosch Frank (2022-12-02 10:03:22) > >> On 12/1/22 09:46, Nico Boehr wrote: > >>> Upcoming changes will add a test which is very similar to the existing > >>> skey migration test. To reduce code duplication, move the common > >>> functions to a library which can be re-used by both tests. > >>> > >> > >> NACK > >> > >> We're not putting test specific code into the library. > > > > What do you mean by "test specific"? After all, it is used by two tests now, possibly more in the future. > > > > Any alternative suggestions? > > For me this is like putting kselftest macros/functions into the kernel. > > The KUT library is more or less the kernel on which the tests in s390x/ > are based on. It provides primitives which (hopefully and mostly) aren't > specific to tests. > > Yes: > Providing skey set and get functions for one or multiple pages to tests. > I.e. sske and iske wrappers. > > No: > Providing multi-page skey set and verify functions that set and verify > skeys based on a pattern which is __hardcoded__ into the function using > the skey wrappers. I.e. you're trying to create a new layer (test > functionality) and stuffing it into the unit test kernel library. > > What you want is a separate testlib which would reside in s390x/testlib/ > where we can store often repeated functions and macros. oufff, then we need to also fix the cmma migration tests