public inbox for ecos-discuss@sourceware.org
 help / color / mirror / Atom feed
* RE: [ECOS] Bug in i386 interrupt handler
@ 2000-10-31 18:19 Fabrice Gautier
  0 siblings, 0 replies; 4+ messages in thread
From: Fabrice Gautier @ 2000-10-31 18:19 UTC (permalink / raw)
  To: 'Phil Hinton', Ecos-List (E-mail)

> -----Original Message-----
> From: Phil Hinton [ mailto:phil@crosstor.com ]
> Subject: [ECOS] Bug in i386 interrupt handler
> 
> Did I miss a patch? Anyone having trouble with i386 PC DSR posts?

I didn't experienced any problem but it really looks like I should have. 

Thanks

-- 
Fabrice Gautier
fabrice_gautier@sdesigns.com 

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [ECOS] Bug in i386 interrupt handler
  2000-10-31 15:25 ` Jonathan Larmour
@ 2000-11-01  8:47   ` Phil Hinton
  0 siblings, 0 replies; 4+ messages in thread
From: Phil Hinton @ 2000-11-01  8:47 UTC (permalink / raw)
  To: Jonathan Larmour; +Cc: Ecos-List (E-mail)

Jonathan Larmour wote:

> Phil Hinton wrote:
> > 
> > The hal_intc_ack macro then uses EAX, and when interrupt_end is
> > called:
> > 
> >              pushl %edx
> >              pushl %eax
> >              call interrupt_end
> > 
> > Instead this should be:
> > 
> >              pushl %edx
> >              pushl %esi   <----  modification
> >              call interrupt_end
> > 
> > Did I miss a patch? Anyone having trouble with i386 PC DSR posts?
> 
> This does indeed look like it would be a problem. If you can just confirm
> that this does fix your problem, I'll update our code here.
> 

The SCSII interrupt handling is now working as expected. I also stepped
through the (assembly level) code to make sure the behavior is as expected.

The problem is fixed.

> Jifl
> -- 
> Red Hat, 35 Cambridge Place, Cambridge, UK. CB2 1NS  Tel: +44 (1223) 728762
> "Plan to be spontaneous tomorrow."  ||  These opinions are all my own fault
> 

Phil Hinton
CrosStor Software, Inc.
908-226-0100 ext 138
phil@crosstor.com



^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [ECOS] Bug in i386 interrupt handler
  2000-10-31 11:54 Phil Hinton
@ 2000-10-31 15:25 ` Jonathan Larmour
  2000-11-01  8:47   ` Phil Hinton
  0 siblings, 1 reply; 4+ messages in thread
From: Jonathan Larmour @ 2000-10-31 15:25 UTC (permalink / raw)
  To: Phil Hinton; +Cc: Ecos-List (E-mail)

Phil Hinton wrote:
> 
> The hal_intc_ack macro then uses EAX, and when interrupt_end is
> called:
> 
>              pushl %edx
>              pushl %eax
>              call interrupt_end
> 
> Instead this should be:
> 
>              pushl %edx
>              pushl %esi   <----  modification
>              call interrupt_end
> 
> Did I miss a patch? Anyone having trouble with i386 PC DSR posts?

This does indeed look like it would be a problem. If you can just confirm
that this does fix your problem, I'll update our code here.

Jifl
-- 
Red Hat, 35 Cambridge Place, Cambridge, UK. CB2 1NS  Tel: +44 (1223) 728762
"Plan to be spontaneous tomorrow."  ||  These opinions are all my own fault

^ permalink raw reply	[flat|nested] 4+ messages in thread

* [ECOS] Bug in i386 interrupt handler
@ 2000-10-31 11:54 Phil Hinton
  2000-10-31 15:25 ` Jonathan Larmour
  0 siblings, 1 reply; 4+ messages in thread
From: Phil Hinton @ 2000-10-31 11:54 UTC (permalink / raw)
  To: Ecos-List (E-mail)

Hi,

I've been trying to get SCSII interrupt drivers working on the i386 PC
eCos using the latest tree from anoncvs, plus various patches from
Fabrice (PCI config, multi-thread, etc).

In our implementation, DSR's were being posted for the SCSII interrupt 
even though the interrupt handler returns CYG_ISR_HANDLED only.

After further investigation I found a bug in hal_pc_irq
(ecos\packages\hal\i386\arch\current\src\vectors.S).

hal_pc_irq calls the interrupt handler, invokes the hal_intc_ack macro,
then calls interrupt_end.

When the interrupt handler returns, EAX has its returned value, which is
to be passed as an argument to interrupt_end. Since the hal_intc_ack 
macro uses the EAX register, it is first stored to ESI by:

            movl %eax, %esi

The hal_intc_ack macro then uses EAX, and when interrupt_end is
called:

             pushl %edx
             pushl %eax
             call interrupt_end

Instead this should be:

             pushl %edx
             pushl %esi   <----  modification
             call interrupt_end

Did I miss a patch? Anyone having trouble with i386 PC DSR posts?

Phil Hinton
CrosStor Software, Inc.
908-226-0100 ext 138
phil@crosstor.com


^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2000-11-01  8:47 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2000-10-31 18:19 [ECOS] Bug in i386 interrupt handler Fabrice Gautier
  -- strict thread matches above, loose matches on Subject: below --
2000-10-31 11:54 Phil Hinton
2000-10-31 15:25 ` Jonathan Larmour
2000-11-01  8:47   ` Phil Hinton

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).