linux-kernel - Re: [PATCH] firmware: arm_scmi: Resolve dependency with TEE subsystem

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Fri, 11 Nov 2022 15:00:29 +0000
From:   Ludvig Pärsson <Ludvig.Parsson@...s.com>
To:     "sumit.garg@...aro.org" <sumit.garg@...aro.org>,
        "sudeep.holla@....com" <sudeep.holla@....com>
CC:     "linux-arm-kernel@...ts.infradead.org" 
        <linux-arm-kernel@...ts.infradead.org>,
        "jens.wiklander@...aro.org" <jens.wiklander@...aro.org>,
        "vincent.guittot@...aro.org" <vincent.guittot@...aro.org>,
        Ludvig Pärsson <Ludvig.Parsson@...s.com>,
        "etienne.carriere@...aro.org" <etienne.carriere@...aro.org>,
        "cristian.marussi@....com" <cristian.marussi@....com>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] firmware: arm_scmi: Resolve dependency with TEE subsystem

On Fri, 2022-11-11 at 14:38 +0000, Sudeep Holla wrote:
> On Fri, Nov 11, 2022 at 03:23:13PM +0530, Sumit Garg wrote:
> > The OP-TEE SCMI transport channel is dependent on TEE subsystem to
> > be
> > initialized first. But currently the Arm SCMI subsystem and TEE
> > subsystem are invoked on the same initcall level as
> > subsystem_init().
> > 
> > It is observed that the SCMI subsystem initcall is invoked prior to
> > TEE
> > subsystem initcall. This leads to unwanted error messages regarding
> > TEE
> > bus is not present yet. Although, -EPROBE_DEFER tries to workaround
> > that
> > problem.
> > 
> > Lets try to resolve inter subsystem dependency problem via shifting
> > Arm
> > SCMI subsystem to subsystem_init_sync() initcall level.
> > 
> 
> I would avoid doing that. We already have some implicit dependency
> with
> subsys_initcall because this driver creates/registers bus and need to
> be
> done early. Now in order to solve the dependency between SCMI and
> TEE,
> both of which creates/registers bus and are at same subsys_initcall,
> we are relying on subsys_initcall_sync.
> 
> Me and Ludvig discussed this in private and I suggested him to do
> something
> like below patch snippet. He mentioned he did post a patch on the
> list but
> I couldn't find it. For this the scmi node must be child node of
> OPTEE as
> it is providing the transport.
> 
> @Ludvig, ?
> 
> Regards,
> Sudeep
> 
> --
> diff --git i/drivers/tee/optee/smc_abi.c
> w/drivers/tee/optee/smc_abi.c
> index a1c1fa1a9c28..839feca0def4 100644
> --- i/drivers/tee/optee/smc_abi.c
> +++ w/drivers/tee/optee/smc_abi.c
> @@ -1534,7 +1534,9 @@ static int optee_probe(struct platform_device
> *pdev)
>                 goto err_disable_shm_cache;
> 
>         pr_info("initialized driver\n");
> -       return 0;
> +
> +       /* Populate any dependent child node(if any) */
> +       return devm_of_platform_populate(&pdev->dev);
> 
>  err_disable_shm_cache:
>         if (!optee->rpc_param_count)
> 
I have answered something similar in my submit [1]. Maybe I should have
CCed you, or atleast sent you this link when I told you I made the
submission.

[1] https://lkml.org/lkml/2022/11/9/803

BR,
Ludvig