From: Bart Veer <bartv@ecoscentric.com>
To: jameshq@hotmail.com
CC: ecos-discuss@sourceware.org
Subject: Re: [ECOS] ref to previous question-'calling path questions,
thanks'
Date: 01 Oct 2006 16:59:13 +0100
>>>>> " " == Qiang Huang <jameshq@hotmail.com> writes:
> Hi eCOS experts:
> Maybe my previous question
>
(http://ecos.sourceware.org/ml/ecos-discuss/2006-09/msg00231.html) is a
> little confusing, what it really means is the followings:
> (ARM target)
> Two thread #1 and #2 exist in system
> (1) Thread #1 runs first
> (2) Thread #1 issues SWI instruction
> (3) Enter SWI ISR and run on thread #1 stack (svc mode)
> (4) At the time IRQ occurs and preempts SWI ISR
> (5) Switch SVC and run on thread #1 stack (svc mode)
> (6) Eventually comes to interrupt_end() and unlock_inner(),
which will cause
> a context switching
> (7) Thread #2 context is loadded and return with thread #2 running
> Here seems SWI is preempted by thread #2? Is this correct?
> (shouldn't all ISR be finished before any thread has the chance
to run?)
> Hope someone can give me some hint, that would be very
> thankful.
eCos does not use SWI's. The main purpose of the SWI instruction is to
provide a controlled mechanism for switching between user mode and
supervisor mode, but eCos applications always run in supervisor mode
so there is no need for a switch. If for some obscure reason an
application does execute a SWI instruction the resulting system
behaviour is basically undefined.