Re: [PATCH v5 7/7] tpm: create TPM 2.0 devices using own device class

[Jarkko Sakkinen <jarkko.sakkinen@linux.intel.com>]

On Tue, Nov 04, 2014 at 01:49:18PM -0700, Jason Gunthorpe wrote:
> On Tue, Nov 04, 2014 at 10:38:43PM +0200, Jarkko Sakkinen wrote:
> 
> > > Then a patch: Drop misc_register entirely, no compat stuff. Explain
> > > clearly the resulting sysfs changes, CC the various people who monitor
> > > the sysfs API, act on any feedback. I'm hoping it is an OK change.
> > > [ If it is not OK then we can talk about using it only for TPM2 or
> > >   whatever ]
> > 
> > Hold on. So you are proposing that for a TPM 1.0 device you would
> > have simultaneously as an intermediate step:
> > 
> > - /sys/class/tpm/tpm0
> > - /sys/class/misc/tpm0
> 
> Yes, well more specifically:
> 
> /sys/class/tpm/tpm0/
> /sys/class/misc/tpm0/dev
> 
> The reason for this is simply that adding to sysfs is certainly OK,
> I don't think the tpm and misc device will collide in any way - udev
> will look for the 'dev' file and ignore the class/tpm directory.
> 
> It breaks up the patch too, this would be a monster:
> 
> > Maybe it would be a better idea to just create patch that would simply
> 
> Because of all the variable rename noise and so forth.
> 
> Even if the net result is we apply several patchs in a row that wipes
> /sys/class/misc/tpm0 the patches themselves will be smaller and more
> reviewable if split, especially those big variable rename ones..

OK, I see your point. I'll break it up into more reasonable steps.

> > Moving the current TPM sysfs attributes is not really an issue because
> > they are only for human consumption (not machine readable in any sany
> > way) :) And AFAIK TrouSerS does not use sysfs PPI interface.
> 
> Someone in user space must use that event log and PPI stuff??? Does
> anyone know who and how?
> 
> Jason

/Jarkko