Re: [tpmdd-devel] [PATCH v1 12/12] tpm: TPM2 sysfs attributes

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

On Tue, Sep 30, 2014 at 02:12:09PM -0600, Jason Gunthorpe wrote:
> On Tue, Sep 30, 2014 at 11:07:21PM +0300, Jarkko Sakkinen wrote:
> > On Fri, Sep 26, 2014 at 08:19:47PM +0300, Jarkko Sakkinen wrote:
> > > On Wed, Sep 24, 2014 at 02:46:27PM -0600, Jason Gunthorpe wrote:
> > > > > That would be 24*2 files only for pcrs...
> > > > 
> > > > Some subsystems do just that..
> > > > 
> > > > $ ls /sys/class/infiniband/qib0/ports/1/sl2vl/
> > > > 0  1  10  11  12  13  14  15  2  3  4  5  6  7  8  9
> > > 
> > > They use static structures in
> > > drivers/infiniband/hw/qib/qib_sysfs.c and it does not looks a
> > > mess. I would prefer to create struct attribute entries
> > > dynamically if there's clean and easy way to do that.
> > 
> > I gave this a shot:
> > 
> > https://github.com/jsakkine/linux-tpm2/commit/dffce68ce34da265a62908dec71b2d85fc16824f
> > 
> > I want to initialize dynamically so that it is easy to support
> > TPM_PT_PCR_COUNT later.
> 
> You can't use a static pcr_dev_attrs, this has to be allocated in the
> chip structure (because of the NULL). Otherwise looks about right
> (although there are more problematic core details here, like racing of the
> tpm dev create with the creation of the sysfs files)

I improved things and pushed a commit that encapsulates PCR bank into
struct pcrs_kobj (internal struct in tpm2-sysfs.c).

This brings me to my question. In TPM2 there is not just PCR bank with
SHA-1 hashes but there are multiple PCR banks.

The way I plan to represent this is:

pcrs/<tag for banks hash algorithm>/<PCR index>

My last commit provides most of the code for implementing this although
in the initial patch set I would take a short cut and only show SHA-1
PCRs. The important thing is that it is put into pcrs/sha1 directory.

> Do you have a reason to have the pcrs in sysfs? I'd be just as happy
> to see them dropped or moved to debugfs for TPM2 as well.

Well, I at least find it useful debugging tool sometimes. And now it
is more useful because PCR values are machine readable.

I will include this to the v2 patch set. When we have a clean patch
set with all the fixes applied into tpm2-v1 branch it will be a better
time to evaluate what is mandatory and what can be postponed.

> Jason

/Jarkko