[ECOS] mbox, ISR and DSR...

[ECOS] mbox, ISR and DSR...

[Fabrice Gautier]

Hi,

Is it safe to use mbox function in an ISR? and in a DSR?

There is no mention of the mbox functions in the table at:
http://sources.redhat.com/ecos/docs-latest/ref/ecos-ref.13.html#pgfId=108056
9


Thanks
-- 
Fabrice Gautier
fabrice_gautier@sdesigns.com 

Re: [ECOS] mbox, ISR and DSR...

[Andrew Lunn]

> Hi,
> 
> Is it safe to use mbox function in an ISR? and in a DSR?

I've used cyg_mbox_tryput() in a DSR. Its worked so far. It seems a
genaral rule of thumb is you can use anything that does not block in a
DSR. 

        Andrew

Re: [ECOS] mbox, ISR and DSR...

[Hugo Tyson]

Andrew Lunn <andrew.lunn@ascom.ch> writes:

> > Hi,
> > 
> > Is it safe to use mbox function in an ISR? and in a DSR?
> 
> I've used cyg_mbox_tryput() in a DSR. Its worked so far. It seems a
> genaral rule of thumb is you can use anything that does not block in a
> DSR. 

Absolutely right.

The reason we have ISRs and DSRs is to allow exactly that:

 o ISRs are called right now regardless of kernel state, therefore you
   cannot do calls into the kernel from an ISR.
 o DSRs are called (when requested) as soon as possible after the
   associated ISR, when the kernel state is such that it's safe to call
   into the kernel.

A DSR still cannot sleep, because it is not a thread context; only threads
can sleep.  Therefore only operations which cannot block may be called from
a DSR.

For example, Andrew says cyg_mbox_tryput() is OK.  Correct.  But
cyg_mbox_put() is NOT OK because it can block if the queue is full.

NB eCos does not (cannot) police this very well; you don't necessarily get
a neat assertion failure if you do call any old thing from an ISR, or if
you call sleep() from a DSR.  Your app will go wrong, that's all; if it
appears to work, it's not working the way you think ;-) Just don't do it.

	- Huge