Re: [PATCH v3] clk: samsung: Prevent potential endless loop in the PLL set_rate ops
From: Sylwester Nawrocki
Date: Thu Sep 17 2020 - 06:21:46 EST
On 15.09.2020 13:34, Sylwester Nawrocki wrote:
> On 14.08.2020 02:46, Chanwoo Choi wrote:
>> On 8/13/20 6:55 PM, Sylwester Nawrocki wrote:
>>> In the .set_rate callback for some PLLs there is a loop polling state
>>> of the PLL lock bit and it may become an endless loop when something
>>> goes wrong with the PLL. For some PLLs there is already code for polling
>>> with a timeout but it uses the ktime API, which doesn't work in some
>>> conditions when the set_rate op is called, in particular during
>>> initialization of the clk provider before the clocksource initialization
>>> has completed. Hence the ktime API cannot be used to reliably detect
>>> the PLL locking timeout.
>>> This patch adds a common helper function for busy waiting on the PLL lock
>>> bit with timeout detection.
>>> Signed-off-by: Sylwester Nawrocki <s.nawrocki@xxxxxxxxxxx>
>>> Changes for v3:
>>> - use busy-loop with udelay() instead of ktime API
>>> Changes for v2:
>>> - use common readl_relaxed_poll_timeout_atomic() macro
>>> drivers/clk/samsung/clk-pll.c | 94 ++++++++++++++++---------------------------
>>> 1 file changed, 34 insertions(+), 60 deletions(-)
>> Acked-by: Chanwoo Choi <cw00.choi@xxxxxxxxxxx>
> Patch applied, thank you for your comments.
And dropped now as it causes issues on arm64. As reported by Marek
it seems udelay() doesn't work before the system timer is initialized.