linux-kernel - Re: [PATCH v5] clk: samsung: Prevent potential endless loop in the PLL ops

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:   Sun, 22 Nov 2020 16:16:02 +0100
From:   Krzysztof Kozlowski <krzk@...nel.org>
To:     Sylwester Nawrocki <s.nawrocki@...sung.com>
Cc:     linux-clk@...r.kernel.org, tomasz.figa@...il.com,
        cw00.choi@...sung.com, m.szyprowski@...sung.com, sboyd@...nel.org,
        mturquette@...libre.com, b.zolnierkie@...sung.com,
        linux-kernel@...r.kernel.org, linux-samsung-soc@...r.kernel.org
Subject: Re: [PATCH v5] clk: samsung: Prevent potential endless loop in the
 PLL ops

On Fri, Nov 20, 2020 at 04:57:31PM +0100, Sylwester Nawrocki wrote:
> The PLL status polling loops in the set_rate callbacks of some PLLs
> have no timeout detection and may become endless loops when something
> goes wrong with the PLL.
> 
> For some PLLs there is already the ktime API based timeout detection,
> but it will not work in all conditions when .set_rate gets called.
> In particular, before the clocksource is initialized or when the
> timekeeping is suspended.
> 
> This patch adds a common helper with the PLL status bit polling and
> timeout detection. For conditions where the timekeeping API should not
> be used a simple readl_relaxed/cpu_relax() busy loop is added with the
> iterations limit derived from measurements of readl_relaxed() execution
> time for various PLL types and Exynos SoCs variants.
> 
> Actual PLL lock time depends on the P divider value, the VCO frequency
> and a constant PLL type specific LOCK_FACTOR and can be calculated as
> 
>  lock_time = Pdiv * LOCK_FACTOR / VCO_freq
> 
> For the ktime API use cases a common timeout value of 20 ms is applied
> for all the PLLs with an assumption that maximum possible value of Pdiv
> is 64, maximum possible LOCK_FACTOR value is 3000 and minimum VCO
> frequency is 24 MHz.
> 
> Signed-off-by: Sylwester Nawrocki <s.nawrocki@...sung.com>

Reviewed-by: Krzysztof Kozlowski <krzk@...nel.org>

Best regards,
Krzysztof