[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: Response time



> -----Original Message-----
> From: owner-dev-etrax@xxxxxxx.com]On">mailto:owner-dev-etrax@xxxxxxx.com]On
> Behalf Of Arne Bockholdt
> Sent: dinsdag 22 oktober 2002 15:56
> To: dev-etrax@xxxxxxx.com
> Subject: Response time
>
>
> Hi !
>
> could anybody on that list make a statement about the min/max/average
> response time for user programs and kernel drivers on the ETRAX LX system
> ? Is there any AXIS supported preemptive patch for the 2.4 kernel or a
> working version of the 2.5 kernel for the ETRAX platform ? Any other
> thoughts about how to perform time critical tasks on the ETRAX system ?

I'm afraid a preemptive patch requires some work, because as far as I
can see the etrax-specific drivers don't provide proper locking
(they assume no preemption in kernel code, so many functions are
not reentrant)

I believe pre-emptive patches make use of the spinlocks that normally
provide
necessary locking for SMP systems.
(in non SMP systems spinlocks are optimised away) Off course it's unlikely
to have an etrax-SMP system in the near future, but this lack of locking
makes applying a preemptive patch a bit more difficult.

I've always been using spinlocks in my kernel code, even if it was for
embedded devices with no change of ever being run on SMP systems, with the
thought of preemptive patches in the back of my head.
And I'm hoping preemption of kernel code will be an option (maybe even a
standard?)
in the future, so I believe all kernel code should have proper locking.


Best regards,

Pieter