[PATCH v6 5/5] optee: add FF-A support

[Sumit Garg]

On Mon, 11 Oct 2021 at 14:59, Jens Wiklander <jens.wiklander at linaro.org> wrote:
>
> On Fri, Oct 8, 2021 at 9:21 PM Jerome Forissier <jerome at forissier.org> wrote:
> >
> > Hi Sudeep,
> >
> > On 10/8/21 6:32 PM, Sudeep Holla wrote:
> > > On Fri, Oct 08, 2021 at 02:27:32PM +0100, Sudeep Holla wrote:
> > >> On Wed, Oct 06, 2021 at 09:09:02AM +0200, Jens Wiklander wrote:
> > >>> Adds support for using FF-A [1] as transport to the OP-TEE driver.
> > >>>
> > >>> Introduces struct optee_msg_param_fmem which carries all information
> > >>> needed when OP-TEE is calling FFA_MEM_RETRIEVE_REQ to get the shared
> > >>> memory reference mapped by the hypervisor in S-EL2. Register usage is
> > >>> also updated to include the information needed.
> > >>>
> > >>> The FF-A part of this driver is enabled if CONFIG_ARM_FFA_TRANSPORT is
> > >>> enabled.
> > >>>
> > >>
> > >> I am not sure if I missed this with earlier version but I see the below
> > >> warning the second time I insert the optee module. I am sure I tested it
> > >> with previous version when I was fixing issues with FF-A as a module.
> > >>
> > >> Not sure if I am missing something in my steps.
> > >>
> > >
> > > OK, more info, indeed I didn't observe this issue 2 weeks back with optee
> > > at commit 57e642f12085 ("core: enable system PTA upon user TA support").
> > > The UUID(d96a5b40-c3e5-21e3-8794-1002a5d5c61b) wasn't enumerated with above
> > > commit. Today I am testing the latest commit 2de2880065f3 ("core: update
> > > reference link to PrimeCell Cache Controller").
> > >
> > > Though it looks like an issue with the driver, the latest optee changes
> > > triggered the driver to hit this issue in my setup. Hope this helps.
> >
> > Totally makes sense. The optee_os commit triggering this is almost
> > certainly 811c42d40cd0 ("plat-vexpress: enable CFG_ENABLE_EMBEDDED_TESTS
> > by default"). With this commit, the "invoke tests" PTA gets included by
> > default in the TEE binary. This PTA has the TA_FLAG_DEVICE_ENUM flag so
> > it is enumerated by the Linux driver. AFAICT the vexpress platform has
> > no other PTA with that flag, so as you said the OP-TEE commit is simply
> >  exposing a bug in the Linux driver.
> >
> > > IIUC we are not undoing optee_register_device executed via
> > > optee_enumerate_devices in the exit path.
> >

Thanks Sudeep for spotting this issue. I have sent a corresponding fix
[1], give it a try and let me know if you observe any further issues.

[1] https://lkml.org/lkml/2021/10/12/136

-Sumit

> > That doesn't sound good indeed ;)
>
> Thanks for the analysis, I'll look into this.
>
> Cheers,
> Jens