On 18-Jan-99 email@example.com wrote:
> On Mon, Jan 18, 1999 at 10:33:59PM +0100, Harald Koerfgen wrote:
>> This is on my DECstation 5000/260 running two "make -j 10 netboot" in
>> parallel on two kernel source trees. I booted in single user mode,
>> inetd manually and opened four telnet sessions :-).
BTW, does that qualify for going multiuser? <grin>
>> The bug was simple but not necessary obvious: duplicate labels in
>> r2300_misc.S and r4k_misc.S which, although not declared as such, were
>> Quick'n'dirty fix: rename those labels.
> Looks like a bug in the assembler or linker. Your patch actually
> make a difference. Can you try to disassemble the object files rsp. the
> kernel executable and find the difference between the good and the
> kernel image?
Well, you were right, Ralf, it doesn't make a difference. A simple test
showed that the behaviour on the serial console line didn't change. Bang
again. At least my test lead me into the right direction: the serial device
drivers was (and still is) in badly need for some polishing.
Applying the attached patch seems to change everything to normal,
including the slowed down init :-(. On the other hand, if such little
changes affect the speed of init I would consider the zs serial device
driver as scrap. Any volunteers? <evil grin>
While speaking about the subject: Tom, what about your DS 2100? Do you see
an abnormal slowdown of init?
P.S.: Please simply forget my patch to r2300_misc.S. This is completly
P.P.S: I am going to do some work on the annoying "Warning: File '...'
has modification time in the future" feature :-).