linux-kernel - Re: [PATCH v3 6/7] arm: dts: sun8i: a83t: Fix undefined offset with virtual timer

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]

Message-ID: <823f256a-3240-7bb4-cc0b-711108f3076c@arm.com>
Date:   Tue, 20 Feb 2018 18:07:54 +0000
From:   Marc Zyngier <marc.zyngier@....com>
To:     Mylène Josserand <mylene.josserand@...tlin.com>,
        maxime.ripard@...tlin.com, wens@...e.org, linux@...linux.org.uk,
        robh+dt@...nel.org, mark.rutland@....com
Cc:     devicetree@...r.kernel.org, quentin.schulz@...tlin.com,
        linux-kernel@...r.kernel.org, clabbe.montjoie@...il.com,
        thomas.petazzoni@...tlin.com, linux-arm-kernel@...ts.infradead.org
Subject: Re: [PATCH v3 6/7] arm: dts: sun8i: a83t: Fix undefined offset with
 virtual timer

Hi Mylène,

On 19/02/18 08:18, Mylène Josserand wrote:
> The ARM architected timers use an offset between their physical and
> virtual counters. That offset should be configured by the bootloader
> in CNTVOFF.
> 
> However, the A83t bootloader fails to do so, and we end up with an
> undefined offset (which in our case is random), meaning that each CPU
> will have a different time, which isn't working very well.
> 
> Fix that by setting the arm,cpu-registers-not-fw-configured that will
> make Linux use the physical timers instead of the virtual ones. One
> possible side effect would be that the virtualization features would
> be disabled. However, due to the way the GIC has been integrated in
> the system, it is already unusable so we're effectively not losing any
> feature.
> 
> Signed-off-by: Mylène Josserand <mylene.josserand@...tlin.com>
> ---
>  arch/arm/boot/dts/sun8i-a83t.dtsi | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/arch/arm/boot/dts/sun8i-a83t.dtsi b/arch/arm/boot/dts/sun8i-a83t.dtsi
> index e97a6d17b8d0..b9bdb891cf2f 100644
> --- a/arch/arm/boot/dts/sun8i-a83t.dtsi
> +++ b/arch/arm/boot/dts/sun8i-a83t.dtsi
> @@ -123,6 +123,7 @@
>  			     <GIC_PPI 14 (GIC_CPU_MASK_SIMPLE(8) | IRQ_TYPE_LEVEL_LOW)>,
>  			     <GIC_PPI 11 (GIC_CPU_MASK_SIMPLE(8) | IRQ_TYPE_LEVEL_LOW)>,
>  			     <GIC_PPI 10 (GIC_CPU_MASK_SIMPLE(8) | IRQ_TYPE_LEVEL_LOW)>;
> +		arm,cpu-registers-not-fw-configured;
>  	};
>  
>  	clocks {
> 

Is the firmware dropping you in the kernel in secure or non-secure mode?

If the later, what you have is the only solution. If the former, that
I'd suggest you adopt what we already have for the Renesas stuff (see
arch/arm/mach-shmobile/headsmp-apmu.S and commit 3fd45a136ff6).

It would allow you to use the virtual timer as intended.

Thanks,

	M.
-- 
Jazz is not dead. It just smells funny...