Subject: Re: [BISECTED] tpm CLKRUN breaks PS/2 keyboard and touchpad on
 Braswell system
To: James Ettle <james@ettle.org.uk>, linux-integrity@vger.kernel.org,
        azhar.shaikh@intel.com, jarkko.sakkinen@linux.intel.com
Cc: linux-kernel@vger.kernel.org, james.l.morris@oracle.com,
        Jason Gunthorpe <jgg@ziepe.ca>
References: <57d96314-cc9c-0656-186e-4eb77a132b70@ettle.org.uk>
From: Javier Martinez Canillas <javierm@redhat.com>
Message-ID: <34b361bf-cce7-a1ac-f8a3-76ef22f5b6b0@redhat.com>
Date: Tue, 12 Dec 2017 12:38:00 +0100
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101
 Thunderbird/52.5.0
MIME-Version: 1.0
In-Reply-To: <57d96314-cc9c-0656-186e-4eb77a132b70@ettle.org.uk>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 2882
Lines: 67

Hello James,

On 12/11/2017 08:37 PM, James Ettle wrote:
> Hello,
> 
> [First: Apologies if cross-posting from Kernel.org BZ is bad form; my distro BZ advised I post this to your mailing list as well.]
> 
> Situation: enabling TPM on a Clevo W510LU with an Intel N3160 CPU breaks PS/2 keyboard and mouse. They just don't respond until after a suspend/resume cycle, and after that they later stop after a while.
> 
> I have confirmed this by blacklisting tpm modules. I noticed this first with kernel 4.13, and have bisected it down to:
> 
> 5e572cab92f0bb56ca1e6e5ee4d807663a7ccbad is the first bad commit
> commit 5e572cab92f0bb56ca1e6e5ee4d807663a7ccbad
> Author: Azhar Shaikh <azhar.shaikh@intel.com>
> Date:   Sun Jun 18 19:17:59 2017 -0700
> 
>     tpm: Enable CLKRUN protocol for Braswell systems
>     
>     To overcome a hardware limitation on Intel Braswell systems,
>     disable CLKRUN protocol during TPM transactions and re-enable
>     once the transaction is completed.
>     
>     Signed-off-by: Azhar Shaikh <azhar.shaikh@intel.com>
>     Reviewed-by: Jarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
>     Tested-by: Jarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
>     Signed-off-by: Jarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
>     Signed-off-by: James Morris <james.l.morris@oracle.com>
> 
> :040000 040000 5437c91886cb62c497255f2c60dbedd7268ab50d 1863a1738ded35a817aad52f9f2b451bd43623d7 M	drivers
> 
> Currently in Kernel.org bugzilla 197287.
> 
> Please let me know if you need any further info.
>

I don't have access to a Braswell machine to reproduce this. I tried to do it
on different Intel systems by modifying is_bsw() to always return true, but
that didn't work either. They work correctly even when CLKRUN_EN is toggled.

I'm not familiar with LPC so please let me know if my assumptions are wrong,
but I find suspicious that a driver for a single device attached to the bus
can control the CLKRUN# signal which AFAIU may be needed for other devices.

So that would explain why the mentioned commit causes issues for PS/2 mouse
and keyboards, since these are attached to the LPC bus and may rely on the
CLKRUN# signal for proper operation.

I see that the following [0,1] patches for the tpm_tis driver landed a few
days ago, they change how the CLKRUN protocol is enabled/disabled. Instead
of doing it per each TPM transaction, it does it once for the duration of
a TPM command.

Not sure if that will make things better or worse for you, but it would be
good to try in case it makes a difference.

[0]: http://git.infradead.org/users/jjs/linux-tpmdd.git/commit/667dcc75be864ff4c17cf58891853b7393bba3e2
[1]: http://git.infradead.org/users/jjs/linux-tpmdd.git/commit/db3248e8a036c39141c8f7e9f1cf5c5ae6815f76

> Many thanks,
> James.
> 

Best regards,
-- 
Javier Martinez Canillas
Software Engineer - Desktop Hardware Enablement
Red Hat