[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <9206176d-2a79-a70d-9b2e-546385b42f3b@canonical.com>
Date: Tue, 19 Oct 2021 18:10:15 +0200
From: Krzysztof Kozlowski <krzysztof.kozlowski@...onical.com>
To: Alexandre Belloni <alexandre.belloni@...tlin.com>,
Sam Protsenko <semen.protsenko@...aro.org>
Cc: Alessandro Zummo <a.zummo@...ertech.it>, linux-rtc@...r.kernel.org,
linux-samsung-soc@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 4/4] rtc: s3c: Fix RTC read on first boot
On 19/10/2021 17:48, Alexandre Belloni wrote:
> On 19/10/2021 16:17:24+0300, Sam Protsenko wrote:
>> On first RTC boot it has the month register value set to 0.
>> Unconditional subtracting of 1 subsequently in s3c_rtc_gettime() leads
>> to the next error message in kernel log:
>>
>> hctosys: unable to read the hardware clock
>>
>> That happens in s3c_rtc_probe() when trying to register the RTC, which
>> in turn tries to read and validate the time. Initialize RTC date/time
>> registers to valid values in probe function on the first boot to prevent
>> such errors.
>>
>
> No, never ever do that, the time is bogus and it has to stay this way,
> else userspace can't know whether the time on the RTC is the actual wall
> time or just some random value that you have set from the driver.
>
Indeed. This looks basically like a revert of your commit 5c78cceeb2d8
("rtc: s3c: stop setting bogus time"). For the Samsung PMIC RTC, we
dropped time initialization in commit fe787a5b2297 ("rtc: s5m: remove
undocumented time init on first boot").
Best regards,
Krzysztof
Powered by blists - more mailing lists