linux-kernel - Re: [PATCH] dell-smm-hwmon: Cache fan_type() calls and use fan

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]

Date:	Mon, 13 Jun 2016 19:02:14 -0700
From:	Guenter Roeck <linux@...ck-us.net>
To:	Pali Rohár <pali.rohar@...il.com>
Cc:	Jan C Peters <jcpeters89@...il.com>,
	Thorsten Leemhuis <fedora@...mhuis.info>,
	David Santamaría Rogado <howl.nsp@...il.com>,
	Peter Saunderson <peteasa@...il.com>,
	Jean Delvare <jdelvare@...e.com>,
	Tolga Cakir <cevelnet@...il.com>, linux-hwmon@...r.kernel.org,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH] dell-smm-hwmon: Cache fan_type() calls and use
 fan_status() for fan detection

On 06/13/2016 11:52 AM, Pali Rohár wrote:
> On Sunday 22 May 2016 02:28:00 Pali Rohár wrote:
>> On Sunday 22 May 2016 02:19:48 Guenter Roeck wrote:
>>> On 05/21/2016 07:46 AM, Pali Rohár wrote:
>>>> On more Dell machines (e.g. Dell Precision M3800)
>>>> I8K_SMM_GET_FAN_TYPE call is too expensive (CPU is too long in
>>>> SMM mode) and cause kernel to hang. This patch cache type for
>>>> each fan (as it should not change) and change the way how fan
>>>> presense is detected. It revert and use function fan_status() as
>>>> was before commit f989e55452c7 ("i8k: Add support for fan
>>>> labels").
>>>>
>>>> Moreover, kernel hangs for 2 - 3 seconds only sometimes and only
>>>> on some Dell machines. When kernel hangs fan speed is at max. So
>>>> it was hard to debug and bisect where is root of this problem.
>>>> It looks like this is bug in Dell BIOS which implement fan type
>>>> SMM code... and there is no way how to fix it in kernel.
>>>>
>>>> Signed-off-by: Pali Rohár <pali.rohar@...il.com>
>>>> Reviewed-by: Jean Delvare <jdelvare@...e.de>
>>>> Reported-and-tested-by: Tolga Cakir <cevelnet@...il.com>
>>>> Fixes: f989e55452c7 ("i8k: Add support for fan labels")
>>>> Link: https://bugzilla.kernel.org/show_bug.cgi?id=112021
>>>> Link: https://bugzilla.kernel.org/show_bug.cgi?id=100121
>>>> Cc: stable@...r.kernel.org # v4.0+, will need backport
>>>
>>> Should this patch be applied, or do you wait for more testing ?
>>
>> I would like to hear some confirmation from people with affected
>> machine.
>>
>
> Ok, now after testing we know that kernel should prevent calling
> I8K_SMM_GET_FAN_TYPE on affected buggy Dell machines.
>
> Looks like there are two different bugs in Dell SMM with
> I8K_SMM_GET_FAN_TYPE call.
>
> First bug cause that kernel freeze for 2 - 3 seconds when
> I8K_SMM_GET_FAN_TYPE is issued.
>
> Second bug cause that fan goes randomly up and down (that is controlled
> by Dell SMM) when I8K_SMM_GET_FAN_TYPE is issued. Normal behaviour is
> returned after machine reboots.
>
> Some Dell machines are affected by first bug, some by second bug. And
> there are Dell machines without both bugs.
>
> This my patch just partially fix first bug and prevent calling that call
> at boot time. But can be issued by sysfs (+value is cached, so it is
> called only once).
>
> So question is: is my patch enough for fixing first bug?
>
> And second question: how to fix second bug? I see only one option:
> Create machine blacklist with broken Dell SMM firmware and disallow
> calling I8K_SMM_GET_FAN_TYPE for them.
>
Or maybe a whitelist with known working systems ?

Guenter