Re: [PATCH 1/3] kasan: support backing vmalloc space with real shadow memory

[Mark Rutland <mark.rutland@arm.com>]

On Thu, Jul 25, 2019 at 12:06:46PM +0200, Marco Elver wrote:
> On Thu, 25 Jul 2019 at 09:51, Dmitry Vyukov <dvyukov@google.com> wrote:
> >
> > On Thu, Jul 25, 2019 at 9:35 AM Dmitry Vyukov <dvyukov@google.com> wrote:
> > >
> > > ,On Thu, Jul 25, 2019 at 7:55 AM Daniel Axtens <dja@axtens.net> wrote:
> > > >
> > > > Hook into vmalloc and vmap, and dynamically allocate real shadow
> > > > memory to back the mappings.
> > > >
> > > > Most mappings in vmalloc space are small, requiring less than a full
> > > > page of shadow space. Allocating a full shadow page per mapping would
> > > > therefore be wasteful. Furthermore, to ensure that different mappings
> > > > use different shadow pages, mappings would have to be aligned to
> > > > KASAN_SHADOW_SCALE_SIZE * PAGE_SIZE.
> > > >
> > > > Instead, share backing space across multiple mappings. Allocate
> > > > a backing page the first time a mapping in vmalloc space uses a
> > > > particular page of the shadow region. Keep this page around
> > > > regardless of whether the mapping is later freed - in the mean time
> > > > the page could have become shared by another vmalloc mapping.
> > > >
> > > > This can in theory lead to unbounded memory growth, but the vmalloc
> > > > allocator is pretty good at reusing addresses, so the practical memory
> > > > usage grows at first but then stays fairly stable.
> > > >
> > > > This requires architecture support to actually use: arches must stop
> > > > mapping the read-only zero page over portion of the shadow region that
> > > > covers the vmalloc space and instead leave it unmapped.
> > > >
> > > > This allows KASAN with VMAP_STACK, and will be needed for architectures
> > > > that do not have a separate module space (e.g. powerpc64, which I am
> > > > currently working on).
> > > >
> > > > Link: https://bugzilla.kernel.org/show_bug.cgi?id=202009
> > > > Signed-off-by: Daniel Axtens <dja@axtens.net>
> > >
> > > Hi Daniel,
> > >
> > > This is awesome! Thanks so much for taking over this!
> > > I agree with memory/simplicity tradeoffs. Provided that virtual
> > > addresses are reused, this should be fine (I hope). If we will ever
> > > need to optimize memory consumption, I would even consider something
> > > like aligning all vmalloc allocations to PAGE_SIZE*KASAN_SHADOW_SCALE
> > > to make things simpler.
> > >
> > > Some comments below.
> >
> > Marco, please test this with your stack overflow test and with
> > syzkaller (to estimate the amount of new OOBs :)). Also are there any
> > concerns with performance/memory consumption for us?
> 
> It appears that stack overflows are *not* detected when KASAN_VMALLOC
> and VMAP_STACK are enabled.
> 
> Tested with:
> insmod drivers/misc/lkdtm/lkdtm.ko cpoint_name=DIRECT cpoint_type=EXHAUST_STACK

Could you elaborate on what exactly happens?

i.e. does the test fail entirely, or is it detected as a fault (but not
reported as a stack overflow)?

If you could post a log, that would be ideal!

Thanks,
Mark.