Bug#542275: kernel error message for #542275 + patch

[Benjamin Cama <benoar@free.fr>]

Le mercredi 03 février 2010 à 02:02 +0000, Ben Hutchings a écrit :
> Error 22 is EINVAL and actually indicates an invalid time.

OK. I actually wanted to solve this bug to also help a bug somewhat
related to #535354, which prevent hwclock to write the hardware clock.

Actually, hwclock first read the hardware clock before writing to it,
and stops if reading fails, which causes my problem (yes, i know, not
this bug).

> If you don't mind applying another patch, this would help debug this
> further:

No problem. But by just reading it, I felt the problem : an invalid date
is stored in the RTC.

What's in my logs (I may have missed the SMU stuff as it seems "make"
didn't feel like recompiling it):

[   20.172405] rtc-generic rtc-generic: rtc core: registered rtc-generic
as rtc0
[   20.186144] pmu_get_time: now = 2454713
[   20.187340] rtc_valid_tm: invalid time { 70, 0, -24077, -14, -8, -7 }
[   20.188559] rtc-generic rtc-generic: hctosys: unable to read the
hardware clock (-22)

So, actually the bug here may be that Jordi simply had his RTC not
correctly set. This often happens with old computers like G3/G4 because
they have worn out batteries and "forget" the date.

But still, it would be usefull to be able to set the RTC for those with
still a bit of juice in their battery. So, this is not this bug but
another one (related to hwclock), if you agree with my analysis. I'll
fill another one soon, if so.

benjamin