Date: Thu, 6 Oct 2016 19:46:21 +0300
From: Jarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
To: Jason Gunthorpe <jgunthorpe@obsidianresearch.com>
Cc: "Winkler, Tomas" <tomas.winkler@intel.com>,
        "tpmdd-devel@lists.sourceforge.net" 
        <tpmdd-devel@lists.sourceforge.net>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH] tpm: don't destroy chip device prematurely
Message-ID: <20161006164621.GA4794@intel.com>
References: <20161002102455.GA27464@intel.com>
 <20161002212126.GA25872@obsidianresearch.com>
 <5B8DA87D05A7694D9FA63FD143655C1B542F466B@hasmsx108.ger.corp.intel.com>
 <20161003124836.GE9990@intel.com>
 <20161004051946.GA10572@intel.com>
 <20161004164738.GA17149@obsidianresearch.com>
 <20161005100234.GA20851@intel.com>
 <20161005162741.GA18636@obsidianresearch.com>
 <20161006112357.GA10533@intel.com>
 <20161006162245.GF1224@obsidianresearch.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20161006162245.GF1224@obsidianresearch.com>
Organization: Intel Finland Oy - BIC 0357606-4 - Westendinkatu 7, 02160 Espoo
User-Agent: Mutt/1.5.24 (2015-08-30)
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1088
Lines: 27

On Thu, Oct 06, 2016 at 10:22:45AM -0600, Jason Gunthorpe wrote:
> On Thu, Oct 06, 2016 at 02:23:57PM +0300, Jarkko Sakkinen wrote:
> 
> > I think that they should be fenced then for the sake of consistency.
> > I do not see why sysfs code is privileged not to do fencing while other
> > peers have to do it.
> 
> Certainly the locking could be changed, but it would be nice to have a
> reason other than aesthetics.
> 
> sysfs is not unique, we also do not grab the rwlock lock during any
> commands executed as part of probe. There are basically two locking
> regimes - stuff that is proven to by synchronous with probe/remove
> (sysfs, probe cmds) and everything else (kapi, cdev)
> 
> Further, the current sysfs implementation is nice and sane: the file
> accesses cannot fail with ENODEV. That is a useful concrete property
> and I don't think we should change it without a good reason.

The last point is certainly legit. I think it even might deserve a
comment of its own in tpm_del_char_device.

I think I have a good idea now what to do. Hold on for RFC patch.

> Jason

/Jarkko