[Click] ELEMENT_MT_SAFE

[Eddie Kohler]

Hey Beyers!

On May 9, 2005, at 6:47 AM, Beyers Cronje wrote:
> On the same note, I have a couple of questions regarding MT click.
>
> If you use StaticThreadSched and configure it such that an element
> will only run on one CPU, is it still necesary to apply locking ?

For interactions between handlers/timers and tasks, yes.  Otherwise no.


> On what CPU will element timers run using StaticThreadSched ? Will it
> run on the same CPU as the timer's element ?

This is not guaranteed (although now that you mention it, it makes 
sense).


> What is the best way to troubleshoot/debug MT click and locking? I'm
> playing around with a custom MT configuration with all elements using
> spinlocks but I'm still getting lockups on Click (1.4.3).

Try newest CVS, which has some locking improvements relative to 1.4.3.


> Is it necessary/recommended to release a lock before calling push() ?

It depends on the type of lock, whether configurations might be 
circular, etc.  I would certainly tend to release the lock before 
pushing, just to minimize the amount of time the lock is held.


> Anyone know of additional documentation on SMP/MT click other than
> http://pdos.csail.mit.edu/papers/click:usenix01/ ?

Unfortunately that's about it for explicit docs....

E



>
> Kind regards
>
> Beyers
>
> _______________________________________________
> click mailing list
> click at amsterdam.lcs.mit.edu
> https://amsterdam.lcs.mit.edu/mailman/listinfo/click