Re: [patch] Real-Time Preemption, -RT-2.6.12-rc2-V0.7.44-00

[Rui Nuno Capela]

> Our first victim!! :-)
>

No kidding!?


> On Tue, 2005-04-05 at 20:06 +0100, Rui Nuno Capela wrote:
>> >
>> I'm having plenty of this on boot, on my SMP/HT desktop (P4/x86), while
>> running RT-V0.7.44-01 (SMP+PREEMPT_RT):
>>
>>   BUG: kstopmachine:xxxx RT task yield()-ing!
>>
>> See sample dmesg and .config on attach.
>>
>> OTOH, on my laptop (P4/UP), all seems to be clear fine.
>>
>
> The kstopmachine is only run on SMP environments, so it won't show up on
> your UP machine.
>
>
>> Is this something to be affraid of? :)
>>
>
> If your box is still running, then no.  But it's now a chore to remove
> the yield from this algorithm, since yields have a possibility to
> deadlock the system.  Although in this particular case, it may not be a
> problem, since the threads that are being waited on are created for
> other CPUs.  But I haven't looked too much into what stop_machine is
> doing so I can very well be wrong.
>
> Thank you for reporting it, we want to weed out all the yields that a RT
> task may call.
>

Now, I'm sure my fears were reasonable. With RT-V0.7.44-02 SMP a can't
even reach an usable system state. This weird BUG: kstopmachine:xxxx RT
task yield()-ing! is flooding the init process to death.

Completely. Hard-reset is the only viable solution to a fast scrolling
console dump, which seems to be in a terrible, nasty and endless loop.

I'm not afraid anymore. RT-0.7.44-02 is useless on my P4 SMP/HT desktop
machine. ;)

Bye.
-- 
rncbc aka Rui Nuno Capela
rncbc@xxxxxxxxx

Attachment: config-2.6.12-rc2-RT-V0.7.44-02.0smp.gz
Description: GNU Zip compressed data