Hi there!

Another question about the real-time clock: one of the 2 board I bought is having a problem with the real-time clock - it is setting the date in 2016 and I promise my car is not a 1982 DeLorean :-)

We tried to set the clock using hwclock under Debian but it did not work (I have not had a chance yet to try it on the other board).

Can anyone confirm how to check the clock?
Thanks in advance
Olivier

There is an RTAS clock driver in /proc/rtas - you can write a value to this and it'll set the clock properly.

The MPC5200B has an RTC built in but it loses power when the board loses power - this is a design flaw of the chip. To fix it Gerald designed in a small PIC with a small embedded flash and a realtime clock (this also allows the ac-back-behaviour and potentially some other power management functionality). So the only way to truly access this PIC is through the RTAS clock.

A little while ago (2.6.9 or thereabouts) a lot of changes were made to in-kernel time code and RTC drivers which means RTAS isn't used in RTC emulation for some reason, and hwclock relies on ISA clocks and the old Linux RTC framework which isn't really the same.

If anyone wants to go ahead and fix this up, that'd be great. The easiest way is to patch the 'hwclock' scripts in your distribution but that means mimicing a lot of the hwclock tool's functionality in script form.

hwclock tool could also be patched to use RTAS.

The other way is to create an RTAS RTC driver in drivers/rtc which uses RTAS.

Take your pick. It is a Linux bug, not a board problem or an RTC problem, that it cannot handle an industry standard abstracted clock interface properly anymore.

_________________
Matt Sealey

Hi Matt!

Thanks for this great, clear report. We'll let you know of any progress on this.

Cheers
Olivier

Hi folks!

Since I created noise here... I may also update you. After swapping the 2 boards we have in Santiago (the second one is at a friend's lab), we swapped the boards... and they both work. It sounds like the problem was the power supply at my friend's: he inherited an old one that worked for other projects but was not fully functional in this case.

The Linux distribution's hwclock utility worked perfectly with a proper power supply.

Thanks for the hints
Cheers
Olivier

Olivier - when are we going to see the pictures and information about your presentation? :) Please update you EFIKA Project blog! Thanks!

R&B :)

_________________
http://bbrv.blogspot.com

I had a simular problem with my Efika. The clock and even the MAC address changed randomly. I found out the that battery dropped dead (around 2V).

Replacing the battery didn't help. Somehow the default values were screwed up and the invalid data didn't allow to set clock. After around a week of trouble I found out how to fix it.

Remove the battery and powerup/powerdown the board (with a proper amount of time between. Let the memory cells settle.) until you can set the clock and it survives a reboot. Then put in in the battery.
IMPORTANT: Put in the battery while the board is powered up and the clock is valid. This ensures the stuff is valid when turning off the board.
Well, your problem sounds exactly like mine. After performing the procedure above all my clock/MAC problems were gone.

Geit

Bill, Raquel, I've updated the blog ( http://www.gautherot.net/efika ). Unfortunately, the folks who were supposed to take pictures never published them :( so I've uploaded the only one I managed to get. I hope you like the type of things we do with this little platform! We should get some more pictures of this week's event, with the Efika-based host, the AVR demo board, its programmer and the USB uplink for host communication - otherwise we'll just take some in the lab and pretend they were taken there ;-) Seriously, we've quite impressed by all the work that can be done. Congratulations!

By the way, shall I take a picture of the Efika in front of a screen with the OpenOffice logo? That's a real eye-catcher... I'll have my full setup back in about 10 days, if you can wait.

Cheers
Olivier