linux-kernel - Re: [PATCH] hwmon: (dell-smm-hwmon) Disable BIOS fan control on SET

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: <97d17b37-0267-ae32-803c-ce8f7a871ab4@debian.org>
Date:   Fri, 15 Nov 2019 14:44:59 +0100
From:   Giovanni Mascellani <gio@...ian.org>
To:     Pali Rohár <pali.rohar@...il.com>
Cc:     Guenter Roeck <linux@...ck-us.net>,
        Jean Delvare <jdelvare@...e.com>, linux-hwmon@...r.kernel.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH] hwmon: (dell-smm-hwmon) Disable BIOS fan control on
 SET_FAN

Hi,

Il 15/11/19 12:29, Pali Rohár ha scritto:
> No. I have not tested that my patch on other models. You can look at
> that my patch link, some people already reported that on some models it
> does not work...

Yes, I saw that. But they are also using other laptops, which could be
excluded by the whitelist until we have a working command for them.

> What is incompatible with Secure Boot? sys_iopl nor sys_ioperm has
> nothing to do with UEFI Secure Boot. They are just ordinary syscalls
> like any other and are executed on kernel side. And IIRC it is up to the
> kernel how it set privileges for I/O ports. Maybe bootloaders under
> Secure Boot can set some other default values, but kernel can overwrite
> them. I really do not see reason why it could be incompatible with UEFI
> Secure Boot nor why it should not work. It just looks like improper
> setup from userspace.

Ah, my fault here: there is a patch to lock down the kernel when it is
started with Secure Boot[1], and I believed that was already in
mainline, but apparently it is not.

 [1]
https://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs.git/commit?id=160a99536afc317b337212dd40eaba341702343e

> It makes sense to have implemented in kernel switching between automatic
> and manual mode as kernel has API for it. But it depends on the fact
> that we know which SMM API to call. And currently it is just some random
> call which we somehow observed that is working on 2 machines and on some
> more other does not. Until we have fully working implementation we
> cannot put it into kernel.

Mmh, but then what is a plausible way forward to have this? Can't we
start populating a whitelist for the machines we already have a solution
for, and add more entries when they are discovered? This would already
give a benefit to the users of supported laptops, without impacting
users of unsupported laptops. My feeling is that if we pretend to have
information for all possible models supported by dell-smm-hwmon, we will
never benefit any user. Or can you suggest a plan?

> What does not make sense for me is to have that "protection" in kernel.

I am not really sure which "protection" you mean. I didn't mean to
introduce any protection from userspace in my patch, I just wanted to
make SET_FAN working. I think that the kernel module cannot (and should
not) reliably protect itself from userspace sending random IO port
reads/writes.

Thanks, Giovanni.
-- 
Giovanni Mascellani <g.mascellani@...il.com>
Postdoc researcher - Université Libre de Bruxelles

Download attachment "signature.asc" of type "application/pgp-signature" (229 bytes)