Re: [RFC][PATCH 1/2] fs proc: make pagemap a privileged interface

[Eric W. Biederman]

Dave Hansen <dave@xxxxxxxx> writes:

> On 03/12/2015 03:35 PM, Andrew Morton wrote:
>> On Mon, 09 Mar 2015 13:43:21 -0700 Dave Hansen <dave@xxxxxxxx> wrote:
>>> From: Dave Hansen <dave.hansen@xxxxxxxxxxxxxxx>
>>>
>>> Physical addresses are sensitive information.  There are
>>> existing, known exploits that are made easier if physical
>>> information is available.  Here is one example:
>>>
>>> 	http://www.cs.columbia.edu/~vpk/papers/ret2dir.sec14.pdf
>> Do we really need to disable pagemap entirely?  What happens if we just
>> obscure the addresses (ie: zero them)?
>
> I think we have 3 basic options:
>
> 1.  Disable it entirely (-EPERM or whatever).  Apps using it break
>     quickly and fairly obviously (diagnosable with an strace)
> 2.  Zero it, or return some nonsensical thing for the physical address
>     portion, but maintain exporting the PTE flags.  Apps only caring
>     about PTE flags work, but anything trying to do lookups in
>     /proc/kpageflags break.  If we zero it, apps pay get confused
>     thinking they have the _actual_ pfn=0.
> 3.  Scramble it in some way obscuring the physical address.  Unscramble
>     it upon access to /proc/kpageflags.
>
> I think you're suggesting (2).  Doesn't that risk silently breaking
> apps?

I think 3 where the scramble is something like AES crypto is likely to
scramble this well and still protect us from plain text attacks. 

>>> pagemap is also the kind of feature that could be used to escalate
>>> privileged from root in to the kernel.  It probably needs to be
>>> protected in the same way that /dev/mem or module loading is in
>>> cases where the kernel needs to be protected from root, thus the
>>> choice to use CAP_SYS_RAWIO.
>> 
>> Confused.  If you have root, you can do mount -o notparanoid.
>
> Good point.  I guess it doesn't protect us much here unless we also
> restrict the ability to remount.

And the ability to unmount... 

A write-once sysctl or a boot time only parameter is much more likely to
be useful in the scenario where you are concerned about root.

Eric

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/