MIME-Version: 1.0
References: <20230130094157.1082712-1-etienne.carriere@linaro.org>
 <20230130094157.1082712-2-etienne.carriere@linaro.org> <CAFA6WYMDP88=3+NnrwTm_u99M-1BZe5ob-bo4mQDRA=QZ9xQXw@mail.gmail.com>
In-Reply-To: <CAFA6WYMDP88=3+NnrwTm_u99M-1BZe5ob-bo4mQDRA=QZ9xQXw@mail.gmail.com>
From:   Jens Wiklander <jens.wiklander@linaro.org>
Date:   Tue, 7 Feb 2023 11:39:43 +0100
Message-ID: <CAHUa44Hkd10Rby1JViiMxC9OcNjU6Fn6Q7Gvm=bCwPDWpiNqSA@mail.gmail.com>
Subject: Re: [PATCH 2/2] firmware: arm_scmi: optee: use optee system invocation
To:     Sumit Garg <sumit.garg@linaro.org>
Cc:     Etienne Carriere <etienne.carriere@linaro.org>,
        linux-kernel@vger.kernel.org, linux-arm-kernel@lists.infradead.org,
        Sudeep Holla <sudeep.holla@arm.com>,
        Cristian Marussi <cristian.marussi@arm.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Tue, Feb 7, 2023 at 10:59 AM Sumit Garg <sumit.garg@linaro.org> wrote:
>
> On Mon, 30 Jan 2023 at 15:12, Etienne Carriere
> <etienne.carriere@linaro.org> wrote:
> >
> > Changes SCMI optee transport to enable sys_service capability of
> > its tee context to leverage provisioned system resources in OP-TEE
> > preventing possible deadlock.
> >
> > Such deadlock could happen when many Linux clients invoke OP-TEE are
> > are all suspended waiting for an OP-TEE RPC request access an SCMI
> > resource through the SCMI OP-TEE PTA service.
> >
> > Signed-off-by: Etienne Carriere <etienne.carriere@linaro.org>
> > ---
> >  drivers/firmware/arm_scmi/optee.c | 3 +++
> >  1 file changed, 3 insertions(+)
> >
> > diff --git a/drivers/firmware/arm_scmi/optee.c b/drivers/firmware/arm_scmi/optee.c
> > index 2a7aeab40e54..91840345e946 100644
> > --- a/drivers/firmware/arm_scmi/optee.c
> > +++ b/drivers/firmware/arm_scmi/optee.c
> > @@ -559,6 +559,9 @@ static int scmi_optee_service_probe(struct device *dev)
> >         if (IS_ERR(tee_ctx))
> >                 return -ENODEV;
> >
> > +       /* SCMI agent can used TEE system service resources */
> > +       tee_ctx->sys_service = true;
> > +
>
> As per the other thread for patch #1, this feature will only be
> available with OP-TEE supporting TEE_GEN_CAP_REG_MEM. Can we add a
> corresponding conditional check here?

What would be gained by that? If a system thread isn't available or
already is busy we're supposed to fall back to normal threads.

Cheers,
Jens

>
> -Sumit
>
> >         agent = devm_kzalloc(dev, sizeof(*agent), GFP_KERNEL);
> >         if (!agent) {
> >                 ret = -ENOMEM;
> > --
> > 2.25.1
> >