Re: SS20 2.4 SMP, irregular boot freezes and swapon segfaults
hi again
> > I get the same drop in bogomips, but I did not get a slower system...
> is the taskswitching and launching (sub)shells in the same speed like
> your 2.2.20-smp kernel? On my sparc a single running task by itself
well, did not notice any special time issues... but...
> was fast but switching between 2 active tasks take neerby 1 sec pause.
... how did you measure that?
or better: what is an active task for you? if i take the meaning of my
computer science lecture an active task is every task that is not
sleeping,
so if such a task switch would take 1 second the system would be sleeping
anyway.
> (may be a spezial Problem on CPU's with BIG! caches like my 142MHz/1M)
> On my 2.2.20-smp system switching is very much! faster.
umm... i've got a dual 125mhz with 1M cache... (511-6326)
so if big cache is the problem i should have it too.
> > > This all seem to be compiler/bintools dependend (gcc 2.95 vs.
> > > egcs32) and some People say, the 2.4.18 and up may be stable on
> > > not-ROSS-cpu's and without smp.
> > Did you try with egcs? I wonder if that is true...
> yes... SuSE say, gcc 2.95.x on sparc is buggy and the only way to get a
> well running 2.2.x Kernel is egcs32. But egcs isn't able to compile
> 2.4.x ImhO gcc 2.95.x is not buggy and able to build a 2.2.x but i'm not
> sure about the 2.4.x Kernel and gcc (code generation) and there are some
> important dependicies not covered by Compiler versions.
well, i have no idea about all this... i just use the standard debian gcc,
which did well until now.
> startupcode/loader in silo and kernel alignment vs.
> bintools Versions vs.
> compiler versions vs.
> Bootprom version (Sparc/Ross) and Number vs.
> CPU Hardware Type of MMU/Core (cpu bugs ?) vs.
> ??? vs. something more i've forgotten to count up?
>
> Take a look in arch dependend kernel source files an you know what i
> mean! It will be a very difficult Task to hunt Bugs in this area...
> damn! And without cpu-docu.... pfff...
so this one could get hard... but it seems the bug has quiet some age by
now and nobody really is concerned about.... except some outsiders.
and the more people report about similar problems the more i feel like
being fooled by the machine, because everyone notices some other edges...
> > > IMHO it is a Problem nearby the Ross MMU or in managing the
> > > CPU-Cache. Can someone post assembler language docu for Ross cpu's
> > > to my Adress? I'll try to hunt the Bug but there is rare docu on the
> > > Web.
> > I asked for this information on sparclinux@vger.kernel.org, but nobody
> > answered... probably they did not like my mail ;)
> > I will try to help if I can, but I'm a greenhorn concerning the kernel
> > code... and sparcs.
> Hm.. guys with sparcs AND linux are the minority in the mainstream
> but guys with sparcs AND linux AND more then 1 ROSS cpu are the crazy
> part of a minority in the mainstream... :-) :-) :-)
> [Greetings to Propellerhead aldo :-)]
umm... sparc && linux && dual-ross gives true for me.
but it won't be the first time i discover myself to be something like an
alien ;)
> I'm a greenhorn too but who cares?
well, i do not ;)
gruess
vom
emi
Reply to: