From mboxrd@z Thu Jan  1 00:00:00 1970
From: gnurou@gmail.com (Alexandre Courbot)
Date: Fri, 14 Jun 2013 17:27:09 +0900
Subject: [PATCH v2 1/3] ARM: tegra: basic support for Trusted Foundations
In-Reply-To: <51BA1B41.6030002@wwwdotorg.org>
References: <1371114745-24710-1-git-send-email-acourbot@nvidia.com>
 <1371114745-24710-2-git-send-email-acourbot@nvidia.com>
 <51BA1B41.6030002@wwwdotorg.org>
Message-ID: <CAAVeFuLc57pV=To5yaE5x9mrVy1yknH2e90QockCiNbEXRm0WQ@mail.gmail.com>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

On Fri, Jun 14, 2013 at 4:19 AM, Stephen Warren <swarren@wwwdotorg.org> wrote:
> On 06/13/2013 03:12 AM, Alexandre Courbot wrote:
>> Add basic support for booting secondary processors on Tegra devices
>> using the Trusted Foundations secure monitor.
>>
>> Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
>> ---
>>  Documentation/devicetree/bindings/arm/tegra.txt    | 11 +++++
>>  .../devicetree/bindings/vendor-prefixes.txt        |  1 +
>>  arch/arm/configs/tegra_defconfig                   |  1 +
>
> The defconfig change should be a separate patch, so that I can squash it
> into any other defconfig updates separately from all the code changes.

Ok, moved that part into its own patch.

>> diff --git a/arch/arm/mach-tegra/firmware.c b/arch/arm/mach-tegra/firmware.c
>
>> +void __init tegra_init_firmware(void)
>> +{
>> +     struct device_node *node;
>> +
>> +     if (!of_have_populated_dt())
>> +             return;
>> +
>> +     node = of_find_compatible_node(NULL, NULL, "tl,trusted-foundations");
>> +     if (node && !IS_ENABLED(CONFIG_TEGRA_TRUSTED_FOUNDATIONS))
>> +             pr_warn("Trusted Foundations detected but support missing!\n");
>> +#if IS_ENABLED(CONFIG_TEGRA_TRUSTED_FOUNDATIONS)
>> +     else if (node)
>> +             register_firmware_ops(&tegra_trusted_foundations_ops);
>> +#endif
>> +}
>
> Is it worth continuing on in the node && !IS_ENABLED case here? After
> all, we can be pretty certain that the write to the CPU reset vector is
> immediately going to trap...

That's what was happening until 3.9, but from 3.10 on the trap is
apparently handled and the boot completes (although with only one
processor).

> I suppose that perhaps without SMP, cpuidle, suspend, ... we could keep
> running, but that seems a little niche.

If we can keep running, even in degraded mode, I see no reason to
panic. The problem is well reported in the kernel log, and having a
running system might be helpful to analyze the issue.

Thanks,
Alex.