[Cbe-oss-dev] RTC problem in PS3

[Geoff Levand]

Hi,

Marvin wrote:
> forward to a better place...
> 
> On Sunday 16 November 2008 23:44:50 Fabiano Manoel de Andrade wrote:
>> 	Hi I've compiled the latest version of linux kernel (2.6.27.6) in my
>> PS3 running debian and get a warning message listed here
>>
>> 	Setting the system clock.
>>   	Cannot access the Hardware Clock via any known method.
>>   	Use the --debug option to see the details of our search for an access
>> method. * Unable to set System Clock to: Sun Nov 16 20:33:06 UTC 2008
>>
>>
>> 	The problem can be reproduced with the commands
>>
>> 	# hwclock --debug
>>  	hwclock from util-linux-ng 2.13.1.1
>>   	hwclock: Open of /dev/rtc failed, errno=2: No such file or directory.
>>   	No usable clock interface found.
>>   	Cannot access the Hardware Clock via any known method.
>>   	# ls -l /dev/rtc*
>>   	ls: cannot access /dev/rtc*: No such file or directory
>>
>> 	Loading the rtc-ppc module solve the problem, but after restart
>> the system I still have the problem. Looking at kernel config I change the
>> rtc-ppc to be build into kernel and this solve totally the problem. So the
>> ps3_defconfig must set the rtc-ppc to be build into kernel.

This is do to a bug in the rtc-ppc driver.  It is not PS3 specific, but
will affect all powerpc machines that use the rtc-ppc driver and have it
built as a loadable module.

Normally, udev will automatically load drivers for the devices present in the
system.  The info udev uses to autoload the driver is not provided by the
rtc-ppc driver, so when it is build as a loadable module udev can't figure
out which driver to load.

Here is some discussion Geert and I had:

-------- Original Message --------
On Tue, 2 Dec 2008, Geert Uytterhoeven wrote:
> On Mon, 1 Dec 2008, Geoff Levand wrote:
> > There is a similar problem when the new generic ppc-rtc driver is
> > built as a module and the startup scripts try to access the RTC.
> 
> Add it to /etc/modules? But then it'll be loaded on all platforms. But as it
> uses the ppc_md infrastructure, that may actually be the right thing to do on
> all PPC platforms.
> 
> However, ppc-rtc looks like a strange driver: it creates a platform_device, but
> doesn't contain the corresponding platform_driver.
> Ah, it creates the platform_device just to be able to pass a struct device to
> rtc_device_register().
> 
> But as all of this is done from module_init(), there's no opportunity for
> autoloading....

It can be solved...

If the creation of the "ppc-rtc" platform device would be moved to builtin
core code (e.g. arch/powerpc/kernel/setup-common.c?), and drivers/rtc/rtc-ppc.c
would become a real platform driver that binds to the "ppc-rtc" platform
device, then it would be autoloaded by udev.

-Geoff