[ECOS] interrupts disabled during DSRs on ARM

[ECOS] interrupts disabled during DSRs on ARM

[Andrew Parlane]

Hi,

Another question regarding the ARM interrupt handling code.

When an IRQ occurs we jump to IRQ: in hal\arm\arch\current\src\vectors.s 
we save the previous state, potentially switch stack, take the scheduler 
lock, find the IRQ number, call the ISR, and then call interrupt_end.

interrupt_end() in kernel\current\src\intr\intr.cxx queues up the DSR if 
asked to, and then unlocks the scheduler. As long as the scheduler 
wasn't locked before the IRQ, this causes the scheduler to call any DSRs 
that are queued up. This happens in call_pending_DSRs_inner() in the 
same file. We disable interrupts, get the pointer to the DSR to call, 
restore interrupts and call the DSR.

In the case I've been describing we are still in the IRQ handler, and as 
such the interrupts are disabled throughout the entirety of this, and 
hence disabled in the DSR.

Looking at: http://ecos.sourceware.org/docs-1.3.1/ref/ecos-ref.c.html in 
the first paragraph it states: "This separation explicitly allows for 
the DSRs to be run with interrupts enabled, thus allowing other 
potentially higher priority interrupts to occur and be processed while 
processing a lower priority interrupt."

Am I missing something here?

Any comments would be welcome.
Andrew Parlane
Carallon ltd.


-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] interrupts disabled during DSRs on ARM

[Nick Garnett]

On 01/11/13 16:30, Andrew Parlane wrote:
> Hi,
> 
> Another question regarding the ARM interrupt handling code.
> 
> When an IRQ occurs we jump to IRQ: in hal\arm\arch\current\src\vectors.s 
> we save the previous state, potentially switch stack, take the scheduler 
> lock, find the IRQ number, call the ISR, and then call interrupt_end.
> 
> interrupt_end() in kernel\current\src\intr\intr.cxx queues up the DSR if 
> asked to, and then unlocks the scheduler. As long as the scheduler 
> wasn't locked before the IRQ, this causes the scheduler to call any DSRs 
> that are queued up. This happens in call_pending_DSRs_inner() in the 
> same file. We disable interrupts, get the pointer to the DSR to call, 
> restore interrupts and call the DSR.
> 
> In the case I've been describing we are still in the IRQ handler, and as 
> such the interrupts are disabled throughout the entirety of this, and 
> hence disabled in the DSR.
> 
> Looking at: http://ecos.sourceware.org/docs-1.3.1/ref/ecos-ref.c.html in 
> the first paragraph it states: "This separation explicitly allows for 
> the DSRs to be run with interrupts enabled, thus allowing other 
> potentially higher priority interrupts to occur and be processed while 
> processing a lower priority interrupt."
> 
> Am I missing something here?

call_pending_DSRs_inner() is called via
cyg_interrupt_call_pending_DSRs() which is itself called by
hal_interrupt_stack_call_pending_DSRs() in vectors.S. This last routine
switches to the interrupt stack and enables interrupts before calling
cyg_interrupt_call_pending_DSRs().


-- 
Nick Garnett                                         Kernel Architect
eCosCentric Limited    http://www.eCosCentric.com    The eCos experts
Barnwell House, Barnwell Drive, Cambridge, UK.   Tel: +44 1223 245571
Registered in England and Wales:                      Reg No: 4422071

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] interrupts disabled during DSRs on ARM

[Andrew Parlane]

Hi,

Thanks for your reply. You are correct. I had 
CYGIMP_HAL_COMMON_INTERRUPTS_USE_INTERRUPT_STACK turned off, which meant 
that the HAL_INTERRUPT_STACK_CALL_PENDING_DSRS macro pointed to 
Cyg_interrupt::call_pending_DSRs_inner() ie. skipping the re-enable 
interrupts bit. Which seems valid in the case you don't have a separate 
interrupt stack.

Thanks for your help,
Andrew
Carallon ltd.

On 01/11/2013 17:11, Nick Garnett wrote:
>
> On 01/11/13 16:30, Andrew Parlane wrote:
>> Hi,
>>
>> Another question regarding the ARM interrupt handling code.
>>
>> When an IRQ occurs we jump to IRQ: in hal\arm\arch\current\src\vectors.s
>> we save the previous state, potentially switch stack, take the scheduler
>> lock, find the IRQ number, call the ISR, and then call interrupt_end.
>>
>> interrupt_end() in kernel\current\src\intr\intr.cxx queues up the DSR if
>> asked to, and then unlocks the scheduler. As long as the scheduler
>> wasn't locked before the IRQ, this causes the scheduler to call any DSRs
>> that are queued up. This happens in call_pending_DSRs_inner() in the
>> same file. We disable interrupts, get the pointer to the DSR to call,
>> restore interrupts and call the DSR.
>>
>> In the case I've been describing we are still in the IRQ handler, and as
>> such the interrupts are disabled throughout the entirety of this, and
>> hence disabled in the DSR.
>>
>> Looking at: http://ecos.sourceware.org/docs-1.3.1/ref/ecos-ref.c.html in
>> the first paragraph it states: "This separation explicitly allows for
>> the DSRs to be run with interrupts enabled, thus allowing other
>> potentially higher priority interrupts to occur and be processed while
>> processing a lower priority interrupt."
>>
>> Am I missing something here?
> call_pending_DSRs_inner() is called via
> cyg_interrupt_call_pending_DSRs() which is itself called by
> hal_interrupt_stack_call_pending_DSRs() in vectors.S. This last routine
> switches to the interrupt stack and enables interrupts before calling
> cyg_interrupt_call_pending_DSRs().
>
>



-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss