Re: [PATCH 0/2][RFC] Try to handle hctosys w/ rtc modules

[Alexander Holler <holler@ahsoftware.de>]

Am 30.06.2014 21:37, schrieb Alessandro Zummo:
> On Sat, 28 Jun 2014 20:54:33 +0200
> Marc Dietrich <marvin24@gmx.de> wrote:
>
>>> Besides that the current hctosys-mechanism doesn't really work with
>>> hot-plugable devices at all. Guess what N will be when you unplug and
>>> plug in such a RTC again.
>>
>> We have a patch in the kernel which binds the rtc number to the
>> hw device, so this even works for hotpluggable devices (at least on
>> systems supporting device-tree). Not sure what your needs are.
>
>   I'm still not sure about how to solve this, maybe it would be better to
>   identify the hctosys device that should be used at boot with
>   a different identifier than rtcX (and move the option to the kernel command line)

Just in case some missunderstood me. I didn't talk about the userland 
but the in-kernel stuff. Currently, it uses the N which was defined by 
in the config (or zero) on resume to update the time. If you look at my 
3. patch, you will see I have changed this there to use the driver (by 
name) which was used at boot to set the time.

Fixing userspace would be easy too by just changing /dev/rtcN to 
/dev/rtc_drivername. But I didn't that in my 3 patches in order to not 
confuse maintainers (and userland people) with too much changes or new 
stuff at once (unsucessfully, but ...).

There still would be a problem about how to handle two RTCs which do use 
the same driver, but I think this is fictitious scenario (and broken if 
that should be for backup purposes) and therefor if someone would really 
design such a system, he should solve this problem himself.

Alexander Holler