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: <b8de6dc0-acfd-3b5f-e57f-9608dc145008@gmail.com>
Date:	Wed, 8 Jun 2016 13:54:35 -0400
From:	"Austin S. Hemmelgarn" <ahferroin7@...il.com>
To:	Guenter Roeck <linux@...ck-us.net>,
	Pali Rohár <pali.rohar@...il.com>
Cc:	Jean Delvare <jdelvare@...e.com>, Mario_Limonciello@...l.com,
	Gabriele Mazzotta <gabriele.mzt@...il.com>,
	Michał Kępień 
	<kernel@...pniu.pl>, linux-hwmon@...r.kernel.org,
	linux-kernel@...r.kernel.org
Subject: Re: dell-smm-hwmon: security problems

On 2016-06-08 13:37, Guenter Roeck wrote:
> On Wed, Jun 08, 2016 at 03:55:48PM +0200, Pali Rohár wrote:
>> On Wednesday 08 June 2016 15:24:10 Guenter Roeck wrote:
>>> On 06/08/2016 02:57 AM, Pali Rohár wrote:
>>>> Hello!
>>>>
>>>> Mario wrote me about two I think security problems in
>>>> dell-smm-hwmon driver and I would like to ask you, how to fix
>>>> them.
>>>>
>>>> 1) File /proc/i8k (exists only when kernel is compiled with
>>>> CONFIG_I8K) exports DMI_PRODUCT_SERIAL and it can be read by
>>>> ordinary user, without root permission. Normally
>>>> DMI_PRODUCT_SERIAL can be read from sysfs file
>>>> /sys/class/dmi/id/product_serial but only by root user.
>>>>
>>>> 2) Via /proc/i8k ordinary user can set fan speed. This is because
>>>> how "restricted" parameter and variable works. Setting fan speed
>>>> by normal non-root user can be dangerous, e.g. malicious
>>>> application under user "nobody" could take control of fans.
>>>>
>>>> Do you have idea how to fix these problems? Just to note that
>>>> /proc/i8k has stable kernel ABI and changing it will break all
>>>> existing i8k* applications. But /proc/i8k is there only for old
>>>> legacy laptops (year 2000).
>>>>
>>>> There is module parameter "restricted" with default value false and
>>>> description: "Allow fan control if SYS_ADMIN capability set".
>>>> Current code do:
>>>>
>>>> 	case I8K_SET_FAN:
>>>> 		if (restricted && !capable(CAP_SYS_ADMIN))
>>>> 			return -EPERM;
>>>>
>>>> For me description is a bit ambiguous. What about setting
>>>> "restricted" by default to true and updating description to
>>>> something like this?
>>>>
>>>> "Disallow fan control when SYS_ADMIN capability is not set
>>>> (default: 1)"
>>>
>>> Sure. I am sure that someone will complain (we learned just recently
>>> that people still use the old commands, after all), but then the old
>>> behavior can be restored by setting the flag to 0.
>>
>> Either setting that flag to 0 or running that tool under root or with
>> capability CAP_SYS_ADMIN.
>>
>>> I would not use a double negative to describe it. Why not just
>>> something like "Allow fan control only if SYS_ADMIN capability set
>>> (default 1)" ?
>>
>> I was thinking about that description too, but there is problem with
>> meaning too...
>>
>> 0 means fan control is allowed for any user
>> 1 means fan control is allowed only for CAP_SYS_ADMIN
>>
>> Description should be unambiguous for situation when flag is set to 0.
>>
> Sorry, I don't understand how a double negation "disallow ... if not set"
> would make things less ambiguous than "allow ... only if set".
Double negatives become ambiguous when you start to deal with the 
possibility of translation or working with people who are not native 
speakers of the language in question.  In English they're traditionally 
considered bad grammar, while in most other languages they are used for 
emphasis and nothing else, and thus are considered by some people to be 
bad form in technical documentation.

Given this particular case, it would probably be the least ambiguous to say:
Restrict fan control to CAP_SYS_ADMIN

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ