Re: [PATCH] UML: add support for KASAN under x86_64

[Patricia Alfonso <trishalfonso@google.com>]

On Mon, Mar 30, 2020 at 1:41 AM Johannes Berg <johannes@sipsolutions.net> wrote:
>
> On Mon, 2020-03-30 at 10:38 +0200, Dmitry Vyukov wrote:
> > On Mon, Mar 30, 2020 at 9:44 AM Johannes Berg <johannes@sipsolutions.net> wrote:
> > > On Fri, 2020-03-20 at 16:18 +0100, Dmitry Vyukov wrote:
> > > > > Wait ... Now you say 0x7fbfffc000, but that is almost fine? I think you
> > > > > confused the values - because I see, on userspace, the following:
> > > >
> > > > Oh, sorry, I copy-pasted wrong number. I meant 0x7fff8000.
> > >
> > > Right, ok.
> > >
> > > > Then I would expect 0x1000 0000 0000 to work, but you say it doesn't...
> > >
> > > So it just occurred to me - as I was mentioning this whole thing to
> > > Richard - that there's probably somewhere some check about whether some
> > > space is userspace or not.
> > >

Yeah, it seems the "Kernel panic - not syncing: Segfault with no mm",
"Kernel mode fault at addr...", and "Kernel tried to access user
memory at addr..." errors all come from segv() in
arch/um/kernel/trap.c due to what I think is this type of check
whether the address is
in userspace or not.

> > > I'm beginning to think that we shouldn't just map this outside of the
> > > kernel memory system, but properly treat it as part of the memory that's
> > > inside. And also use KASAN_VMALLOC.
> > >
> > > We can probably still have it at 0x7fff8000, just need to make sure we
> > > actually map it? I tried with vm_area_add_early() but it didn't really
> > > work once you have vmalloc() stuff...
> >

What x86 does when KASAN_VMALLOC is disabled is make all vmalloc
region accesses succeed by default
by using the early shadow memory to have completely unpoisoned and
unpoisonable read-only pages for all of vmalloc (which includes
modules). When KASAN_VMALLOC is enabled in x86, the shadow memory is not
allocated for the vmalloc region at startup. New chunks of shadow
memory are allocated and unpoisoned every time there's a vmalloc()
call. A similar thing might have to be done here by mprotect()ing
the vmalloc space as read only, unpoisoned without KASAN_VMALLOC. This
issue here is that
kasan_init runs so early in the process that the vmalloc region for
uml is not setup yet.


> > But we do mmap it, no? See kasan_init() -> kasan_map_memory() -> mmap.
>
> Of course. But I meant inside the UML PTE system. We end up *unmapping*
> it when loading modules, because it overlaps vmalloc space, and then we
> vfree() something again, and unmap it ... because of the overlap.
>
> And if it's *not* in the vmalloc area, then the kernel doesn't consider
> it valid, and we seem to often just fault when trying to determine
> whether it's valid kernel memory or not ... Though I'm not really sure I
> understand the failure part of this case well yet.
>

I have been testing this issue in a multitude of ways and have only
been getting more confused. It's still very unclear where exactly the
problem occurs, mostly because the errors I found most frequently were
reported in segv(), but the stack traces never contained segv.

Does anyone know if/how UML determines if memory being accessed is
kernel or user memory?

> johannes
>


--
Best,
Patricia