linux-kernel - Re: [rtc-linux] [PATCH] drivers/rtc/rtc-pl031.c: reset registers in init flow

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] [day] [month] [year] [list]

Message-ID: <CACRpkdYnr8bw_3a7jnaSZqMmA6j+yyVKNrdA+TzSTdNxzaUu+A@mail.gmail.com>
Date:	Mon, 3 Aug 2015 10:01:31 +0200
From:	Linus Walleij <linus.walleij@...aro.org>
To:	"rtc-linux@...glegroups.com" <rtc-linux@...glegroups.com>
Cc:	Alessandro Zummo <a.zummo@...ertech.it>,
	Alexandre Belloni <alexandre.belloni@...e-electrons.com>,
	"linux-arm-kernel@...ts.infradead.org" 
	<linux-arm-kernel@...ts.infradead.org>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	Deepak Saxena <dsaxena@...xity.net>
Subject: Re: [rtc-linux] [PATCH] drivers/rtc/rtc-pl031.c: reset registers in
 init flow

On Wed, Jul 29, 2015 at 1:17 PM, Leo Yan <leo.yan@...aro.org> wrote:

> This issue is found when i use file
> kernel/power/suspend_test.c to verify the system suspend. Before the
> system suspend, the flow need set 10's alarm in RTC for waken up
> event.
>
> But what i observed the phenomenon is:
>
> At the init phase:
> RTC_LR = 0x0;
> RTC_DR = 0x0;
> RTC_MR = 0x0;
>
> According to the timeout is 10s, the function *pl031_set_alarm()* will
> set RTC_MR = 10; So usually RTC_DR will increase one for every second
> and will trigger interrupt when RTC_DR equal to RTC_MR. But on
> Hi6220, though the RTC_DR init value is zero, but very soon it will be
> set a random value which is bigger than 10.

Aha, that's annoying.

> So it will never match with
> RTC_MR register anymore; finally the system cannot resume back due the
> waken up event will not be triggered.

OK I see.

> So suspect RTC_LR has not been initailized correctly and it will load
> random value to RTC_DR. After reset the RTC_LR, this issue will be
> fixed.
>
>> > +       /* Init registers */
>> > +       writel(0x0, ldata->base + RTC_LR);
>>
>> This will reset the clock to jan 1st 1970 on every reboot.
>> The idea is that the RTC should *preserve* the system time
>> if you reboot the system, so NACK.
>>
>> Usually userspace has a script using hwclock to read the
>> system time from the rtc to system time with hwclock -s
>> after userspace comes up. Likewise it writes it back with
>> hwclock -w before rebooting.
>
> This change is wrong.

I don't see what you mean here...

Most RTCs in the world have a battery back-up, so they
sustain time during shutdown. On an ARM system like this,
this PL031 derivative probably loose the time on shutdown,
but not on a soft reset.

However probe() will be called no matter if a shutdown or
soft reset happened, and the time will be reset to
1970-01-01 in any case with this change, even if it was
a soft reset and the time in the RTC is actually valid.

>> > +       writel(RTC_BIT_AI, ldata->base + RTC_ICR);
>>
>> So why do we want to have the alarm enabled by
>> default, before the kernel nor userspace has requested
>> it?
>
> This is to clear any pending interrupt, but not enable alarm.

OK.

>
>> If your problem is with suspend/resume I suggest you work
>> on the [runtime]_suspend/resume hooks instead of probe().
>> Possibly you need to save/restore state across suspend/resume.
>
> Do you think below change is make sense?

It's not using suspend/resume hooks but let's see...

> +       /* Init registers */
> +       writel(0x0, ldata->base + RTC_IMSC);
> +       writel(RTC_BIT_AI, ldata->base + RTC_ICR);

Looks OK.

> @@ -368,6 +372,9 @@ static int pl031_probe(struct amba_device *adev, const struct amba_id *id)
>                                 writel(time, ldata->base + RTC_LR);
>                         }
>                 }
> +       } else {
> +               time = readl(ldata->base + RTC_DR);
> +               writel(time, ldata->base + RTC_LR);
>         }

This badly needs a comment in the else-clause describing
what is happening here.

But I think it looks right! This will preserve the time across
soft reboots properly if I read it right.

Yours,
Linus Walleij
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/