[Click] Task update: SMP test?

[Eddie Kohler]

Hi Joonwoo, all,

I hope your laptop has been restored to health!!  Very sorry to hear 
about that!

So I implemented user-level multithreading mostly to try to replicate 
this bug.  I am not able to reproduce it.  However, in the mean time, 
some changes to Task were made that could have fixed the underlying problem.

Can anyone using SMP Click verify that it is broken, or report that it 
works for them?

Eddie


Joonwoo Park wrote:
> 2008/5/22 Eddie Kohler <kohler at cs.ucla.edu>:
>> Hi all,
>>
>> I have just checked in some changes to Click's Task layer that should
>> improve robustness in limited cases -- more Task functions may be called
>> from interrupt context than previously.  I would appreciate that SMP Click
>> users test the current Git sources, in case I've broken anything.
>>
>> Thanks,
>> Eddie
>>
> 
> Hi Eddie,
> I've tested it. But I think there is a ploblem on recent code.
> I tested it with some config for stress test, and I had worked pretty
> well for multi-threaded script.
> But when I make tasks to switching their home_thread, it occured panic.
> I tested it on my laptop which is running 2.6.19.2 x86_64.
> 
> I hope that I remember my config correctly:
> in0::InfiniteSource -> Queue
> -> EtherEncap(0x0800, FF:FF:FF:FF:FF:FF, FF:FF:FF:FF:FF:FF)
> -> td0::ToDevice(eth0)
> fd0::FromDevice(eth1) -> Queue -> td1::ToDevice(eth1)
> // BalancedThreadSched(INTERVAL 1)  // makes panic
> // StaticThreadSched(in0 0, td0 1, fd0 0, td11) // works fine
> 
> p.s I screwed up my laptop while digging this problem by spilling a
> ton of water on it. Moreover since I am in Vancouver Canada now, it
> was my only computer.
> :-(:-(:-(:-(:-(:-(:-(:-(
> I'll try to debug it again, when my laptop is repaired.
> I'm sorry for insufficient information.
> 
> Thanks,
> Joonwoo