On Mon, Jan 25, 1999 at 11:17:24PM +0100, Harald Koerfgen wrote:
> On 25-Jan-99 Thomas Riemer wrote:
> > 1. I'm getting errors that my Makefiles have modifications in the
> > future.
> > 2. I'm getting errors that make is detecting a clock skew.
> > 3. I do not have CONFIG_RTC set in the kernel.
> > 4. /sbin/clock doesn't exist.
> > Somehow I think all of these are related to some basic problem with
> > time in the kernel? Does this stuff work on someone elses machine?
> Yes, it does on mine :-). Well, not really, the patch isn't ready for
> rollout yet.
> The cause is that the DECstation timer tick is running at 128 Hz and all
> the userland binaries think that the kernel is running at 100 Hz, which is
> the standard for Linux except for Alpha machines. BTW, DECstation-Linux
> 2.1.121 was running at 64 HZ.
> So we have two possibilities here:
> One is to recompile all userland binaries with HZ=128, thus not beeing able
> to use, for example, the Cobaltmicro rpms. Not very good.
> The second one is to pretend to userland, the kernel is running at 100 Hz.
> The kernel itself is prepared for HZ != 100, otherwise it would not be
> working on Alphas. The culprit are functions where the jiffies or
> derivatives of that are exported to userland. I am working on this.
> Fellow DECstation hackers, please do *not* use CONFIG_RTC. The RTC is the
> only possible source for periodic interrupts on the DECstations and the
> RTC driver likes to play around with that causing bad things to happen. We
> have to disable some features in drivers/char/rtc.c before it's safe to
> use it.
Which binaries do we talk about ? Currently i am not aware which binaries
in Userland deal with the HZ value.
Where is the difficulty in letting the Mips kernel also run at 100Hz ?
128 vs 100 is dropping every fifth (nearly :) )...
Florian Lohoff email@example.com +49-5241-470566
Good, Fast, Cheap: Pick any two (you can't have all three). (RFC 1925)