I have created an lkm that simply calls 'cpu_initclocks()'.
It is a syscall lkm which, when called, fixes the stopped clock on my
system.  Not that time is right any longer, mind you, but at least I
can function.

	Now that the test cycle for stopping the clock is shorter, I
think that my symptoms are due to serial activity (not netscape as
previously stated --- at least the two are related).  While the matter
seems exsaserbated now that I have 24M ram (as opposed to 16), I
believe that this is due to the fact that excessive swapping kept
serial performance down for the most part.

	If nobody else experiences this, then I have a slight
suspicion.  I have increased my ZLRB_RING_SIZE to 4096 (and the other
to 4095) which gives me much greater serial performance.  I'm thinking
that somehow this isn't allocated correctly ... or ... there's some
interaction with this larger size and things around it.

	So... who else has a 4/260 to help track this down?

Dave.

-- 
----------------------------------------------------------------------------
|David Gilbert, PCI, Richmond Hill, Ontario.  | Two things can only be     |
|Mail:      dgilbert@pci.on.ca                |  equal if and only if they |
|http://www.pci.on.ca/~dgilbert               |   are precisely opposite.  |
---------------------------------------------------------GLO----------------