linux-kernel - Re: [PATCH] usercopy: Skip multi-page bounds checking on SLOB

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <1471530118.2581.13.camel@redhat.com>
Date:	Thu, 18 Aug 2016 10:21:58 -0400
From:	Rik van Riel <riel@...hat.com>
To:	Kees Cook <keescook@...omium.org>,
	Linus Torvalds <torvalds@...ux-foundation.org>
Cc:	Laura Abbott <labbott@...oraproject.org>,
	linux-kernel@...r.kernel.org, linux-mm@...ck.org,
	xiaolong.ye@...el.com
Subject: Re: [PATCH] usercopy: Skip multi-page bounds checking on SLOB

On Wed, 2016-08-17 at 15:29 -0700, Kees Cook wrote:
> When an allocator does not mark all allocations as PageSlab, or does
> not
> mark multipage allocations with __GFP_COMP, hardened usercopy cannot
> correctly validate the allocation. SLOB lacks this, so short-circuit
> the checking for the allocators that aren't marked with
> CONFIG_HAVE_HARDENED_USERCOPY_ALLOCATOR. This also updates the config
> help and corrects a typo in the usercopy comments.
> 
> Reported-by: xiaolong.ye@...el.com
> Signed-off-by: Kees Cook <keescook@...omium.org>

There may still be some subsystems that do not
go through kmalloc for multi-page allocations,
and also do not use __GFP_COMP

I do not know whether there are, but if they exist
those would still trip up the same way SLOB got
tripped up before your patch.

One big question I have for Linus is, do we want
to allow code that does a higher order allocation,
and then frees part of it in smaller orders, or
individual pages, and keeps using the remainder?

>From both a hardening and a simple stability
point of view, allowing memory to be allocated
in one size, and freed in another, seems like
it could be asking for bugs.

If we decide we do not want to allow that,
we can just do the __GFP_COMP markings
unconditionally, and show a big fat warning
when memory gets freed in a different size
than it was allocated.

Is that something we want to do?