Attaching userspace VM to kernel thread (was Re: [PATCH] fs/binfmt_elf: Fix memsz > filesz handling)

[Kees Cook <keescook@chromium.org>]

Hi,

This has diverged from the original topic a bit, so I've changed the
Subject to hopefully gain visibility. :)

For KUnit, it would be REALLY nice to have a way to attach a userspace
VM to a kernel thread so we can do userspace memory mapping
manipulation, etc. Neither David nor I have been able to figure out the
right set of steps to make this happen. What are we missing?

Details below...

On Wed, Nov 16, 2022 at 12:34:40PM +0800, David Gow wrote:
> On Mon, Nov 7, 2022 at 11:59 AM Kees Cook <keescook@chromium.org> wrote:
> >
> > On Sun, Nov 06, 2022 at 02:16:57AM +0000, Pedro Falcato wrote:
> > David, has there been any work on adding a way to instantiate
> > userspace VMAs in a KUnit test? I tried to write this myself, but I
> > couldn't figure out how to make the userspace memory mappings appear.
> > Here's my fumbling attempt:
> > https://git.kernel.org/pub/scm/linux/kernel/git/kees/linux.git/log/?h=devel/kunit/usercopy
> >
> > I really wish KUnit had userspace mapping support -- I have a bunch of
> > unit tests that need to get built up around checking for regressions
> > here, etc.
> 
> Hi Kees,
> 
> Sorry the the delayed response!
> 
> Alas, my attempts to get this to work haven't been much more
> successful than yours. It's definitely something we'd like to support,
> but I confess to not knowing enough about the mm code to know exactly
> what would be involved.
> 
> The workaround is to load tests as modules, and use something like
> Vitor's original patch here:
> https://lore.kernel.org/all/20200721174036.71072-1-vitor@massaru.org/
> 
> Basically, using the existing mm of the module loader. Adapting those
> changes to your branch (and fixing a couple of back-to-front KUnit
> assertions) does work for me when built as a module, in an x86_64 vm:
> 
> root@slicestar:~# modprobe usercopy_kunit
> [   52.986290]     # Subtest: usercopy
> [   52.986701]     1..1
> [   53.246058]     ok 1 - usercopy_test
> [   53.246628] ok 1 - usercopy
> 
> But getting it to work with built-in tests hasn't been successful so
> far. I wondered if we could just piggy-back on init_mm or similar, but
> that doesn't seem to work either.
> 
> So, in the short-term, this is only possible for modules. If that's
> useful enough, we can get Vitor's support patch (or something similar)
> in, and just mark any tests module-only (or have them skip if there's
> no mm). Because kunit.py only runs built-in tests, though, it's
> definitely less convenient.

Thanks for any pointers! :)

-Kees

-- 
Kees Cook