linux-kernel - Re: [PATCH 2/2] ramoops: remove module parameters

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: <CAGXu5jKHaP4GsuSh9P=+3QJEWhaAy1F_RghuQJdSzmfh0wNAiQ@mail.gmail.com>
Date:	Tue, 22 Nov 2011 10:14:48 -0800
From:	Kees Cook <keescook@...omium.org>
To:	Marco Stornelli <marco.stornelli@...il.com>
Cc:	linux-kernel@...r.kernel.org,
	Chen Gong <gong.chen@...ux.intel.com>,
	Greg Kroah-Hartman <gregkh@...e.de>,
	Andrew Morton <akpm@...ux-foundation.org>,
	Arnd Bergmann <arnd@...db.de>,
	Nicolas Pitre <nicolas.pitre@...aro.org>,
	Paul Gortmaker <paul.gortmaker@...driver.com>
Subject: Re: [PATCH 2/2] ramoops: remove module parameters

On Tue, Nov 22, 2011 at 9:23 AM, Marco Stornelli
<marco.stornelli@...il.com> wrote:
> Il 21/11/2011 19:11, Kees Cook ha scritto:
>>
>> On Sat, Nov 19, 2011 at 1:25 AM, Marco Stornelli
>> <marco.stornelli@...il.com>  wrote:
>>>
>>> Il 18/11/2011 20:31, Kees Cook ha scritto:
>>>>
>>>> The ramoops driver is intended to be used with platforms that define
>>>> persistent memory regions. If memory regions were configurable with
>>>> module parameters, it would be possible to read some RAM regions via
>>>> the pstore interface without access to /dev/mem (which would result
>>>> in a loss of kernel memory privacy when a system is built with
>>>> STRICT_DEVMEM), so remove this ability completely.
>>>>
>>>
>>> I don't like it very much. The loss of module parameters give us less
>>> flexibility. The main goal of this driver is debug, so I think it should
>>> be
>>> fast to use. I mean it's not more possible reserve a memory region and
>>> load
>>> the module "on-the-fly", it needs a platform device, it's ok but I think
>>> it's a little bit more complicated, (without talking about platforms
>>> without
>>> a device tree source).
>>> I don't understand the problem of strict devmem. We shouldn't use kernel
>>> memory region but only reserved ones and the driver doesn't use the
>>> request_mem_region_exclusive, am I wrong?
>>
>> Hmmm, maybe I'm reading it backwards, but I think we want it to use
>> ..._exclusive().
>>
>> int devmem_is_allowed(unsigned long pagenr)
>> {
>>         if (pagenr<= 256)
>>                 return 1;
>>         if (iomem_is_exclusive(pagenr<<  PAGE_SHIFT))
>>                 return 0;
>>         if (!page_is_ram(pagenr))
>>                 return 1;
>>         return 0;
>> }
>>
>> If the region is exclusive, access is not allowed (return 0). ramoops
>> currently uses request_mem_region() instead of
>> request_mem_region_exclusive(). If we made that switch, I think I'd be
>> happy. Would this create some problem I'm not seeing?
>
> I don't understand why we should use the exclusive version, to protect debug
> data? You should provide a more valid reason to change, because the fact you
> will be happier with this change is not enough for me :)

I guess ..._exclusive() doesn't matter. My concern was that ramoops
with the pstore interface and the module parameters could be used to
bypass STRICT_DEVMEM if it were able to be loaded in some sensitive
region of system memory. Perhaps the better approach would be to use a
magic header so that uninitialized memory isn't visible? What do you
think?

-Kees

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