Re: [PATCH v3] tpm_tis: Add missing tpm_request/relinquish_locality calls

From: Jarkko Sakkinen <jarkko@kernel.org>
To: Lukasz Majczak <lma@semihalf.com>
Cc: Dirk Gouders <dirk@gouders.net>,
	Guenter Roeck <linux@roeck-us.net>,
	James Bottomley <James.Bottomley@hansenpartnership.com>,
	Tj <ml.linux@elloe.vision>, Peter Huewe <peterhuewe@gmx.de>,
	Jason Gunthorpe <jgg@ziepe.ca>,
	linux-integrity@vger.kernel.org, linux-kernel@vger.kernel.org,
	Radoslaw Biernacki <rad@semihalf.com>,
	Marcin Wojtas <mw@semihalf.com>, Alex Levin <levinale@google.com>
Subject: Re: [PATCH v3] tpm_tis: Add missing tpm_request/relinquish_locality calls
Date: Thu, 4 Feb 2021 01:24:04 +0200	[thread overview]
Message-ID: <YBswlPuVINENIoUC@kernel.org> (raw)
In-Reply-To: <CAFJ_xbpv3O5f4MPUhZExA3g-4NjPN_vvv7q6Srh1ddqCH4ie9w@mail.gmail.com>

On Wed, Feb 03, 2021 at 02:43:44PM +0100, Lukasz Majczak wrote:
> Hi Jarkko, Dirk,
> 
> Jarkko,
> Thank you for your points - I will try to fix all you have mentioned.
> I think it would be good to clarify a few things, before going with
> the next version. Regarding use tpm_chip_start/stop() around
> tpm2_probe() call - I have followed the similar way it is done in the
> probe_itpm() function, where is also a call to tpm_tis_send_data()
> guarded by request/release_locality(). I have tested it on the Samsung
> Chromebook Pro (which reports TPM 1.2 / Cr50) and it was sufficient
> (e.g. I didn't have to enable the clock) to get rid of a trace
> mentioned in the commit message....but now writing these words I'm
> starting to think that using tpm_chip_start/stop() could be safer from
> a point of view of other TPMs... so if you suggest using
> tpm_chip_start/stop() shall I also add it to the probe_itpm() (instead
> of  request/release_locality()) ?
> 
> Dirk,
> Thanks for the clarification. Regarding the issue you observe - I
> wanted to address at first the one that is generating the trace
> (please check the commit message) because it was leading to returning
> an error in the tpm_tis_status() function causing TPM module not
> initialized at all in the end - requesting locality before the call to
> the tpm_tis_send_data() has helped in my case (my test environment). I
> am aware of the second issue - "TPM interrupt not working, polling
> instead", but as it is not as critical as the first one, I decided to
> work on it later.
> 
> Thank you once again for all your input and sorry for a confusion with
> sending patches.

NP, thanks for doing this. Just take your time and polish the cosmetic
things. Especially for bug fixes a clean changelog is essential.

> Best regards,
> Lukasz

/Jarkko