On Fri, Feb 25, 2022 at 05:46:57PM -0800, Andrew Morton wrote:
> On Fri, 25 Feb 2022 17:35:49 -0800 Kees Cook <keescook@xxxxxxxxxxxx> wrote:
>
> > On Fri, Feb 25, 2022 at 04:01:57PM -0800, Andrew Morton wrote:
> > > On Fri, 25 Feb 2022 09:33:45 -0800 Kees Cook <keescook@xxxxxxxxxxxx> wrote:
> > >
> > > > Under CONFIG_HARDENED_USERCOPY=y, when exact stack frame boundary checking
> > > > is not available (i.e. everything except x86 with FRAME_POINTER), check
> > > > a stack object as being at least "current depth valid", in the sense
> > > > that any object within the stack region but not between start-of-stack
> > > > and current_stack_pointer should be considered unavailable (i.e. its
> > > > lifetime is from a call no longer present on the stack).
> > > >
> > > > Introduce ARCH_HAS_CURRENT_STACK_POINTER to track which architectures
> > > > have actually implemented the common global register alias.
> > > >
> > > > Additionally report usercopy bounds checking failures with an offset
> > > > from current_stack_pointer, which may assist with diagnosing failures.
> > > >
> > > > The LKDTM USERCOPY_STACK_FRAME_TO and USERCOPY_STACK_FRAME_FROM tests
> > > > (once slightly adjusted in a separate patch) will pass again with
> > > > this fixed.
> > >
> > > Again, what does this actually do?
> >
> > [answers]
> >
>
> OK, thanks. I think a new changelog is warranted?
Yup, I've cut/pasted most of that into the new changelog:
usercopy: Check valid lifetime via stack depth
One of the things that CONFIG_HARDENED_USERCOPY sanity-checks is whether
an object that is about to be copied to/from userspace is overlapping
the stack at all. If it is, it performs a number of inexpensive
bounds checks. One of the finer-grained checks is whether an object
crosses stack frames within the stack region. Doing this on x86 with
CONFIG_FRAME_POINTER was cheap/easy. Doing it with ORC was deemed too
heavy, and was left out (a while ago), leaving the courser whole-stack
check.
The LKDTM tests USERCOPY_STACK_FRAME_TO and USERCOPY_STACK_FRAME_FROM
try to exercise these cross-frame cases to validate the defense is
working. They have been failing ever since ORC was added (which was
expected). While Muhammad was investigating various LKDTM failures[1],
he asked me for additional details on them, and I realized that when
exact stack frame boundary checking is not available (i.e. everything
except x86 with FRAME_POINTER), it could check if a stack object is at
least "current depth valid", in the sense that any object within the
stack region but not between start-of-stack and current_stack_pointer
should be considered unavailable (i.e. its lifetime is from a call no
longer present on the stack).
Introduce ARCH_HAS_CURRENT_STACK_POINTER to track which architectures
have actually implemented the common global register alias.
Additionally report usercopy bounds checking failures with an offset
from current_stack_pointer, which may assist with diagnosing failures.
The LKDTM USERCOPY_STACK_FRAME_TO and USERCOPY_STACK_FRAME_FROM tests
(once slightly adjusted in a separate patch) pass again with this fixed.
[1] https://github.com/kernelci/kernelci-project/issues/84
Cc: Matthew Wilcox (Oracle) <willy@xxxxxxxxxxxxx>
Cc: Josh Poimboeuf <jpoimboe@xxxxxxxxxx>
Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
Cc: linux-mm@xxxxxxxxx
Reported-by: Muhammad Usama Anjum <usama.anjum@xxxxxxxxxxxxx>
Signed-off-by: Kees Cook <keescook@xxxxxxxxxxxx>
---
v1: https://lore.kernel.org/lkml/20220216201449.2087956-1-keescook@xxxxxxxxxxxx
v2: https://lore.kernel.org/lkml/20220224060342.1855457-1-keescook@xxxxxxxxxxxx
v3: https://lore.kernel.org/lkml/20220225173345.3358109-1-keescook@xxxxxxxxxxxx
v4: - improve commit log (akpm)
> What's your preferred path for upstreaming this change?
I figured I would take it via my for-next/hardening tree; I have 2
arch changes ready to go (Acked by maintainers) there too (to add
current_stack_pointer).
Thanks for the review!
--
Kees Cook
