Re: [PATCH][next] rtc: msc313: Fix unintentional sign extension issue on left shift of a u16

[Romain Perier]

Hi,

Le mar. 28 sept. 2021 à 15:31, Daniel Palmer <daniel@xxxxxxxx> a écrit :
>
> Hi Colin,
>
> On Tue, 28 Sept 2021 at 21:39, Colin King <colin.king@xxxxxxxxxxxxx> wrote:
> >Shifting the u16 value returned by readw by 16 bits to the left
> >will be promoted to a 32 bit signed int and then sign-extended
> >to an unsigned long. If the top bit of the readw is set then
> >the shifted value will be sign extended and the top 32 bits of
> >the result will be set.

Good catch !

>
> Ah,.. C is fun in all the wrong places. :)
> These chips are full of 32bit registers that are split into two 16
> registers 4 bytes apart when seen from the ARM CPU so we probably have
> this same mistake in a few other places.
>
> A similar pattern is used a bit later on in the same file to read the counter:
>
> seconds = readw(priv->rtc_base + REG_RTC_CNT_VAL_L)
> | (readw(priv->rtc_base + REG_RTC_CNT_VAL_H) << 16);
>
> I guess it works at the moment because the top bit won't be set until 2038.

The crazy stuff being, I ran rtctest from selftests and rtc-range (1)
that tests a variety
of dates including 2038 and 2106 for example. Both tests passed :) (probably
because *this case* specifically did not happen while running the test)

1. https://git.kernel.org/pub/scm/linux/kernel/git/abelloni/rtc-tools.git/tree/rtc-range.c

Thanks,
Regards,
Romain