[U-Boot] [PATCH 00/18] Introduce SPI TPM v2.0 support

[Miquel Raynal]

Hi Simon and Tom,

On Fri, 23 Mar 2018 08:42:02 -0600, Simon Glass <sjg at chromium.org>
wrote:

> Hi Miquel,
> 
> On 21 March 2018 at 07:49, Tom Rini <trini at konsulko.com> wrote:
> > On Tue, Mar 20, 2018 at 03:51:22PM +0100, Miquel Raynal wrote:  
> >> Hi Tom,
> >>
> >> On Tue, 20 Mar 2018 10:04:55 -0400, Tom Rini <trini at konsulko.com> wrote:
> >>  
> >> > On Tue, Mar 20, 2018 at 02:36:56PM +0100, Miquel Raynal wrote:  
> >> > > Hi Tom,
> >> > >
> >> > > Sorry for the delay.
> >> > >
> >> > > On Fri, 9 Mar 2018 07:18:40 -0500, Tom Rini <trini at konsulko.com> wrote:
> >> > >  
> >> > > > On Fri, Mar 09, 2018 at 08:53:40AM +0100, Miquel Raynal wrote:  
> >> > > > > Hi Tom,
> >> > > > >
> >> > > > > On Thu, 8 Mar 2018 12:20:30 -0500, Tom Rini <trini at konsulko.com> wrote:
> >> > > > >  
> >> > > > > > On Thu, Mar 08, 2018 at 04:40:03PM +0100, Miquel Raynal wrote:
> >> > > > > >  
> >> > > > > > > Current U-Boot supports TPM v1.2 specification. The new specification
> >> > > > > > > (v2.0) is not backward compatible and renames/introduces several
> >> > > > > > > functions.
> >> > > > > > >
> >> > > > > > > This series introduces a new SPI driver following the TPM v2.0
> >> > > > > > > specification. It has been tested on a ST TPM but should be usable with
> >> > > > > > > others v2.0 compliant chips.
> >> > > > > > >
> >> > > > > > > Then, basic functionalities are introduced one by one for the v2.0
> >> > > > > > > specification. The INIT command now can receive a parameter to
> >> > > > > > > distinguish further TPMv1/TPMv2 commands. After that, the library itself
> >> > > > > > > will know which one is pertinent and will return a special error if the
> >> > > > > > > desired command is not supported for the selected specification.  
> >> > > > > >
> >> > > > > > Thanks for doing all of this.  Can you please enable this feature on
> >> > > > > > sandbox and/or an x86 QEMU variant where I assume we could also then
> >> > > > > > setup automated testing?
> >> > > > > >  
> >> > > > >
> >> > > > > Not sure I understand your request correctly: the TPM commands are
> >> > > > > already available in the sandbox (I don't see what I could add), I just
> >> > > > > extended the current set of commands.
> >> > > > >
> >> > > > > However, even with these commands, we won't be able to test them in a
> >> > > > > sandbox unless with an actual device.
> >> > > > >
> >> > > > > I probably miss something, can you explain a bit more what you would
> >> > > > > like?  
> >> > > >
> >> > > > Can we add a valid TPM via QEMU and then test it that way?  If so, we
> >> > > > should enable the TPM code on qemu-x86_64 (and, well, if we can pass it
> >> > > > on other arches, other QEMU targets) and write some test/py/tests/ code
> >> > > > that exercises the TPM commands.  Does that make sense?
> >> > > >  
> >> > >
> >> > > I suppose this is doable, but for what I know, the effort is
> >> > > consequent. TPM 2.0 are not compatible at all with TPM 1.x , the
> >> > > packets exchanged at TPM level are completely different. Hence, I
> >> > > think there is almost nothing that we can take from the TPM 1.x
> >> > > implementation already existing in QEMU.  
> >> >
> >> > Ah, OK.  I thought QEMU had a TPM 2.0 implementation now too, but I see
> >> > I'm mistaken.
> >> >  
> >> > > I am certain we all would benefit such a contribution, however I'm
> >> > > not sure I could handle that anytime soon.
> >> > >
> >> > > About the series, I think it would be better that I change a macro name
> >> > > ("STRINGIFY", which is wrongly named), I will send a v2 soon, can you
> >> > > tell me its status otherwise?  
> >> >
> >> > We have the usual linux/stringify.h header available, so yes, you should
> >> > make use of that.  
> >>
> >> Actually the name is misleading as I don't want to "stringify".
> >>
> >> I am looking for a way to easily fill a buffer of bytes from integer
> >> values, ie:
> >>
> >> u32 value = 0x12345678;
> >> u8 buf[x] = { MACRO(value), ...} to be {0x12, 0x34, 0x56, 0x78, ...}
> >>
> >>  
> >> >  And I still would like to see tests written, even if
> >> > they can only be executed on $board with $TPM attached via $interface,
> >> > with those 3 variables documented so that others can try it out too.
> >> > Does that make sense?  Thanks!  
> >>
> >> I see some TPM tests for v1.x, I can probably add some there. This will
> >> test the library functions but not the "user" commands.
> >>
> >> To test the commands, I suggest following the lines I inserted in my
> >> cover letter, but maybe I can put it also in some documentation?
> >>
> >> Would this fit your expectations?  
> >
> > We have a framework to run those commands on the target and confirm that
> > they behave as expected.  Please write the tests to run those commands
> > and confirm that they work as expected.  Thanks!  
> 
> Re sandbox, it has a TPM emulator used for testing in
> tpm_tis_sandbox.c - you should be able to add something similar for
> v2.

I wrote a complete test suite for all the TPMv2.0 commands. It runs
completely fine with real hardware, see the last patch of the v2 series.
I had a look to sandbox but I'm sorry, I really don't have enough time
for now to achieve this additional work, I really hope the full Python
test suite will be enough.

Regards,
Miquèl

-- 
Miquel Raynal, Bootlin (formerly Free Electrons)
Embedded Linux and Kernel engineering
https://bootlin.com