New Sheevaplugs - less bogomips - timebase 10% too fast :-(

[docbee]

A reseller of my software got a bunch of new SheevaPlugs that just show 992 BogoMips instead of 1192 my older ones. Both are running a 2.6.29 debian image. What makes things even more concerning is, that the clock of the new ones runs about 10% too fast, ntpd reports enormous jitter to time servers when inspected with "ntpq -p" while pings are going out flawlessly.

WHAT HAS CHANGED WITH THE NEW SHEEVAPLUGS IN REGARD TO CLOCK?

[docbee]

I did compare "dmesg" on old and new SheevaPlug:

old: [42949373.220000] Kirkwood: MV88F6281-A0, TCLK=200000000.
new: [42949373.220000] Kirkwood: MV88F6281-Rev-Unsupported, TCLK=166666667.

Looks like they changed some system chips without further notice! 

[guidol]

new: [42949373.220000] Kirkwood: MV88F6281-Rev-Unsupported, TCLK=166666667.
Looks like they changed some system chips without further notice! 

The "MV88F6281-Rev-Unsupported" seems to be a "MV88F6281-A1" which looks like
not be to so good supported in your kernel version 2.6.29.

On the following link it is suggested to use 2.6.32 which will include a patch to
support the new revision. Take a look at the following URL:
http://www.mail-archive.com/debian-arm@lists.debian.org/msg10506.html

[tbm]

2.6.29 is not a kernel Debian ships, so it's probably a vendor kernel.  Since they ship a kernel that doesn't work on their hardware, you should complain to them.

However, it would be nice if you could confirm that Debian's 2.6.32 kernel works.

[docbee]

@guidol: thanks for the hint with 2.6.32.

My complaining is about GlobalScale changing silicon without giving notice and bringing people trying to do business with sheevaplug into a very difficult situation in regard to their customer base.

@tbm: I don't complain about debian. I really love the work you and others have done to port debian onto the sheeva. When I know that things change I can adapt to this before sending systems to customers and to find out the hard way, that something has changed with the hardware.

However, 1000 thanks for helping me to get out of this very unpleasant situation.

I will report if the 2.6.32-kirkwood kernel does fix the issue.

[docbee]

good news, 2.6.32 does the trick.