Re: [PATCHv2 2/2] tpm_ftpm_tee: register driver on TEE bus

[Sumit Garg]

+ Thiru (original author), Ilias, Joakim

Hi Jarkko,

On Mon, 25 May 2020 at 12:20, Maxim Uvarov <maxim.uvarov@xxxxxxxxxx> wrote:
>
> On Fri, 22 May 2020 at 23:03, Jarkko Sakkinen
> <jarkko.sakkinen@xxxxxxxxxxxxxxx> wrote:
> >
> > On Fri, May 22, 2020 at 10:29:44PM +0300, Maxim Uvarov wrote:
> > > On Fri, 22 May 2020 at 20:15, Jarkko Sakkinen
> > > <jarkko.sakkinen@xxxxxxxxxxxxxxx> wrote:
> > > >
> > > > On Thu, May 21, 2020 at 09:47:43AM +0300, Maxim Uvarov wrote:
> > > > > Register driver on TEE bus. module tee registers bus,
> > > >
> > > > "on the TEE bus"
> > > >
> > > > "The module tee"
> > > >
> > > > > and module optee calls optee_enumerate_devices() to scan
> > > > > all devices on the bus. Trusted Application for this driver
> > > >
> > > > Looking at drivers/tee, it shows that tee and optee are in fact the same
> > > > module as opposed to what your commit message says.
> > > >
> > >
> > > In the current kernel it's 2 different modules.
> > >
> > > > > can be Early TA's (can be compiled into optee-os). In that
> > > > > case it will be on OPTEE bus before linux booting. Also
> > > > > optee-suplicant application is needed to be loaded between
> > > > > OPTEE module and ftpm module to maintain functionality
> > > > > for fTPM driver.
> > > >
> > > > Why is this needed and why things worked before having this?

If you remembered earlier discussions when this driver was added,
tee-supplicant dependency concern was raised at that time too. But the
response from author [1] was to use a workaround in firmware to
overcome the initialization issue of fTPM driver.

Basically while using upstream OP-TEE, fTPM NV RAM is implemented via
RPMB secure storage [2]. So any fTPM operation dependent on NV RAM may
fail without access to RPMB secure storage.

And during kernel boot, RPMB access isn't available (there were some
prior efforts [3] around this but unfortunately didn't land in
upstream) due to which we need to rely on user mode tee-supplicant for
corresponding access.

So we should only register fTPM device when the underlying
infrastructure is working, which is something this patch-set addresses
via registering fTPM device only when the tee-supplicant is up and
running.

[1] https://lkml.org/lkml/2019/7/4/779
[2] https://optee.readthedocs.io/en/latest/architecture/secure_storage.html#rpmb-secure-storage
[3] https://lwn.net/Articles/682276/

> > > >
> > >
> > > Before these changes user space has to drive boot sequence. I.e.  tee
> > > and optee modules loads, then application tee-supplicant has to start
> > > and only then module ftpm can be loaded. The reason for that is
> > > storage services are implemented in userspace and driver needs them.
> >
> > Is the TPM implementation uploaded to TEE from user space and or what
> > storage are we talking about? Not sure how these storage services
> > connect to the TPM.

Please see my response above.

-Sumit

> >
> > /Jarkko
>
> Jakko,
> tee-supplicant application provides state machine over callbacks with
> RPC messages.
> https://github.com/OP-TEE/optee_client/blob/master/tee-supplicant/src/tee_supplicant.c#L614
> It also allocates shm. Without running tee-supplicant
> tee_client_open_session() will fail.
> optee_open_session()->get_msg_arg()->tee_shm_alloc()->...
> Optee team wanted to remove some dependencies from tee-supplicant with
> moving code
> to the kernel. But for now I think that should be out of the scope of
> current patches due to
> they fix driver initialization on tee bus without breaking current
> functionality.
>
> Maxim.