On Fri, 19 Mar 2021 at 11:46, Vlastimil Babka <vbabka@xxxxxxx> wrote: > On 3/18/21 12:47 PM, Marco Elver wrote: > > On Tue, Mar 16, 2021 at 01:41PM +0100, glittao@xxxxxxxxx wrote: > >> From: Oliver Glitta <glittao@xxxxxxxxx> > >> > >> SLUB has resiliency_test() function which is hidden behind #ifdef > >> SLUB_RESILIENCY_TEST that is not part of Kconfig, so nobody > >> runs it. Kselftest should proper replacement for it. > >> > >> Try changing byte in redzone after allocation and changing > >> pointer to next free node, first byte, 50th byte and redzone > >> byte. Check if validation finds errors. > >> > >> There are several differences from the original resiliency test: > >> Tests create own caches with known state instead of corrupting > >> shared kmalloc caches. > >> > >> The corruption of freepointer uses correct offset, the original > >> resiliency test got broken with freepointer changes. > >> > >> Scratch changing random byte test, because it does not have > >> meaning in this form where we need deterministic results. > >> > >> Add new option CONFIG_TEST_SLUB in Kconfig. > >> > >> Add parameter to function validate_slab_cache() to return > >> number of errors in cache. > >> > >> Signed-off-by: Oliver Glitta <glittao@xxxxxxxxx> > > > > No objection per-se, but have you considered a KUnit-based test instead? > > To be honest, we didn't realize about that option. > > > There is no user space portion required to run this test, and a pure > > in-kernel KUnit test would be cleaner. Various boiler-plate below, > > including pr_err()s, the kselftest script etc. would simply not be > > necessary. > > > > This is only a suggestion, but just want to make sure you've considered > > the option and weighed its pros/cons. > > Thanks for the suggestion. But I hope we would expand the tests later to e.g. > check the contents of various SLUB related sysfs files or even write to them, > and for that goal kselftest seems to be a better starting place? Not sure, but I would probably go about it this way: A. Anything that is purely in-kernel and doesn't require a user space component should be a KUnit test. B. For any test that requires a user space component, it'd be a kselftest. And I think the best design here would also clearly separate those 2 types of tests, and I wouldn't lump tests of type A into modules that are also used for B. That way, running tests of type A also is a bit easier, and if somebody wants to just quickly run those it's e.g. very quick to do so with kunit-tool. Thanks, -- Marco
