[Adeos-main] [PATCH 0/2] Fasteoi fixes

[Adeos-main] [PATCH 0/2] Fasteoi fixes

[Wolfgang Mauerer]

Dear all,

following Jan's considerations, I found that we lack a generic
bug fix for the fasteoi handler in 2.6.38, which is backported
in the first patch. The second patch makes ipipe agree with
threaded one-shot handlers.

Pending adaption to the latest API, this part is also relevant
for the ipipe-3.0-noarch. Philippe, do you agree with the
approach?

Cheers, Wolfgang

--
Siemens AG, Open Source Platforms,
Corporate Competence Centre Embedded Linux

[Adeos-main] [PATCH 1/2] Fix fasteoi masking

[Wolfgang Mauerer]

To prevent interrupt storms, the fasteoi handler must mask
the interrupt line in oneshot mode. Backported from c69e3758ff56.
---
 kernel/irq/chip.c |    3 +++
 1 files changed, 3 insertions(+), 0 deletions(-)

diff --git a/kernel/irq/chip.c b/kernel/irq/chip.c
index 9671529..2d25c62 100644
--- a/kernel/irq/chip.c
+++ b/kernel/irq/chip.c
@@ -575,6 +575,9 @@ handle_fasteoi_irq(unsigned int irq, struct irq_desc *desc)
 		goto out;
 	}
 
+	if (desc->status & IRQ_ONESHOT)
+		mask_irq(desc);
+
 	desc->status |= IRQ_INPROGRESS;
 	desc->status &= ~IRQ_PENDING;
 	raw_spin_unlock(&desc->lock);
-- 
1.7.6.4

[Adeos-main] [PATCH 2/2] Don't unconditionally unmask fasteoi interrupts

[Wolfgang Mauerer]

For threaded interrupt handlers in oneshot mode, the
hard irq handler will not yet have satisfied the device,
so we need to keep the line masked until irq_finalize_oneshot().
---
 kernel/irq/chip.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)

diff --git a/kernel/irq/chip.c b/kernel/irq/chip.c
index 2d25c62..b79ec1d 100644
--- a/kernel/irq/chip.c
+++ b/kernel/irq/chip.c
@@ -589,7 +589,8 @@ handle_fasteoi_irq(unsigned int irq, struct irq_desc *desc)
 	raw_spin_lock(&desc->lock);
 	desc->status &= ~IRQ_INPROGRESS;
 #ifdef CONFIG_IPIPE
-	desc->irq_data.chip->irq_unmask(&desc->irq_data);
+	if (!(desc->status & IRQ_MASKED))
+		desc->irq_data.chip->irq_unmask(&desc->irq_data);
 out:
 #else
 out:
-- 
1.7.6.4

Re: [Adeos-main] [PATCH 0/2] Fasteoi fixes

[Philippe Gerum]

On 11/30/2011 12:58 PM, Wolfgang Mauerer wrote:
> Dear all,
>
> following Jan's considerations, I found that we lack a generic
> bug fix for the fasteoi handler in 2.6.38, which is backported
> in the first patch. The second patch makes ipipe agree with
> threaded one-shot handlers.
>
> Pending adaption to the latest API, this part is also relevant
> for the ipipe-3.0-noarch. Philippe, do you agree with the
> approach?
>

Fine with me. Merged, thanks.

> Cheers, Wolfgang
>
> --
> Siemens AG, Open Source Platforms,
> Corporate Competence Centre Embedded Linux
>
>


-- 
Philippe.