Skip to contentSkip to conversation

genirq: irq_chip->startup() usage in setup_irq and set_irq_chained handler

PATCH In review Pawel MOLL 3 participants 8 messages Aug 21, 2008 – Sep 6, 2008 1 message not in the archive IRQ

lore
Share & tools
lore

Pawel MOLL

author
#0 lore raw

PATCH 1/1 genirq: irq_chip->startup() usage in setup_irq and set_irq_chained handler

This patch clarifies a usage of irq_chip->startup() callback:

1. The "if (startup) startup(); else enabled();" code in setup_irq() is unnecessary, as startup() falls back to enabled() via default callbacks, set by irq_chip_set_defaults().

2. When using set_irq_chained_handler() the startup() was never called, which is not good at all... Fixed. And again - when startup() is not defined the call will fall back to enable() than to unmask() via default callbacks.

---
 kernel/irq/chip.c   |    2 +-
 kernel/irq/manage.c |    5 +----
 2 files changed, 2 insertions(+), 5 deletions(-)
kernel/irq/chip.c +1 -1
index 7279484..a32c337 100644
@@ -595,7 +595,7 @@ __set_irq_handler(unsigned int irq, irq_flow_handler_t handle, int is_chained,
 		desc->status &= ~IRQ_DISABLED;
 		desc->status |= IRQ_NOREQUEST | IRQ_NOPROBE;
 		desc->depth = 0;
-		desc->chip->unmask(irq);
+		desc->chip->startup(irq);
 	}
 	spin_unlock_irqrestore(&desc->lock, flags);
 }
kernel/irq/manage.c +1 -4
index fdccfd5..d0af30c 100644
@@ -369,10 +369,7 @@ int setup_irq(unsigned int irq, struct irqaction *new)
 		if (!(desc->status & IRQ_NOAUTOEN)) {
 			desc->depth = 0;
 			desc->status &= ~IRQ_DISABLED;
-			if (desc->chip->startup)
-				desc->chip->startup(irq);
-			else
-				desc->chip->enable(irq);
+			desc->chip->startup(irq);
 		} else
 			/* Undo nested disables: */
 			desc->depth = 1;
Signature · 1 line
1.5.5.1
author
#1 lore raw

Hi, Ingo,

Apparently I have swapped To: and Cc: field in the original mail, apologies.

Quoted 10 lines

This patch clarifies a usage of irq_chip->startup() callback:

1. The "if (startup) startup(); else enabled();" code in setup_irq() is unnecessary, as startup() falls back to enabled() via default callbacks, set by irq_chip_set_defaults().

2. When using set_irq_chained_handler() the startup() was never called, which is not good at all... Fixed. And again - when startup() is not defined the call will fall back to enable() than to unmask() via default callbacks.

Any comment on the patch?

Best regards

Paweł

#2 lore raw
Quoted 11 lines · Pawel MOLL

Hi, Ingo,

Apparently I have swapped To: and Cc: field in the original mail, apologies.

Quoted 6 lines

This patch clarifies a usage of irq_chip->startup() callback:

1. The "if (startup) startup(); else enabled();" code in setup_irq() is unnecessary, as startup() falls back to enabled() via default callbacks, set by irq_chip_set_defaults().

Quoted 6 lines
Quoted 4 lines

2. When using set_irq_chained_handler() the startup() was never called, which is not good at all... Fixed. And again - when startup() is not defined the call will fall back to enable() than to unmask() via default callbacks.

Any comment on the patch?

looks good to me at first glance - but i Cc:-ed Thomas and Ben, maybe they have further comments.

Ingo

#3 lore raw
Quoted 14 lines · Ingo Molnar
Quoted 14 lines

Apparently I have swapped To: and Cc: field in the original mail, apologies.

Quoted 10 lines

This patch clarifies a usage of irq_chip->startup() callback:

1. The "if (startup) startup(); else enabled();" code in setup_irq() is unnecessary, as startup() falls back to enabled() via default callbacks, set by irq_chip_set_defaults().

2. When using set_irq_chained_handler() the startup() was never called, which is not good at all... Fixed. And again - when startup() is not defined the call will fall back to enable() than to unmask() via default callbacks.

Quoted 4 lines
Quoted 1 line

Any comment on the patch?

looks good to me at first glance - but i Cc:-ed Thomas and Ben, maybe they have further comments.

The second change is a significant semantic change. I wouldn't be surprised if I have cases that rely (or work around) the lack of startup() in set_irq_chained_handler(). I'll have to dbl check things next week. Can you send me a copy of those patches ?

Cheers,
Ben.

author
#4 lore raw
Quoted 4 lines

The second change is a significant semantic change. I wouldn't be surprised if I have cases that rely (or work around) the lack of startup() in set_irq_chained_handler(). I'll have to dbl check things next week.

Let me briefly explain my situation. I have a main interrupt controller which provides startup() and unmask/mask() functions. The first one is rather expensive (as the controller itself is... hmmm... complicated ;-), the second - very cheap. And that is how I understand the different "levels" of interrupt access - startup() should be called once, somewhere during request_irq(), (un)masking may be used frequently.

And one of the interrupt is generated by hardware PIO controller. The idea was obvious - register a chained handler, which decodes the PIO controller state and generates a interrupt, which number may be obtained by gpio_to_irq(). Sounds simple, doesn't it? :-)

And in that moment the problem raised its ugly head - the interrupt controller's startup() was never called for the PIO interrupt (as there was no request_irq()), so the hardware wasn't configured properly and... well... bad things were happening ;-)

So unless I totally misunderstood the meaning of irq_chip callbacks, I believe the startup() should be called in set_irq_chained_handler().

Regards

Paweł

#5 lore raw
Quoted 7 lines · Pawel MOLL

Let me briefly explain my situation. I have a main interrupt controller which provides startup() and unmask/mask() functions. The first one is rather expensive (as the controller itself is... hmmm... complicated ;-), the second - very cheap. And that is how I understand the different "levels" of interrupt access - startup() should be called once, somewhere during request_irq(), (un)masking may be used frequently.

Oh, I don't disagree. It's probably a good idea. I'm just worried of the potential impact on existing code written around the current behaviour.

We have 23 calls to set_irq_chained_handler in arch/powerpc, and I need to audit them all. Luckily, we mostly don't have startup() callbacks.

(... some time later ...)

It looks good. Of course, we'll have to test at one point, but at this stage, I think powerpc is happy with the change.

Interestingly enough, I can see a case where we would have a problem -without- your change :-) Not with the current code, but in conjunction with another change that's planned for .28.

So Acked-by: Benjamin Herrenschmidt <[email protected]>

Cheers,
Ben.

author
#6 lore raw
Quoted 1 line

So Acked-by: Benjamin Herrenschmidt <[email protected]>

Great, thanks ;-)

Ingo, I will send you an acked patch.

Cheers

Paweł

A message that is not in the archive (sent off-list or not yet received): [email protected]

Acked-by Benjamin Herrenschmidt
#8 lore raw
Quoted 6 lines · Pawel MOLL

This patch clarifies usage of irq_chip->startup() callback:

1. The "if (startup) startup(); else enabled();" code in setup_irq() is unnecessary, as startup() falls back to enabled() via default callbacks, set by irq_chip_set_defaults().

Quoted 7 lines

2. When using set_irq_chained_handler() the startup() was never called, which is not good at all... Fixed. And again - when startup() is not defined the call will fall back to enable() than to unmask() via default callbacks.

applied the commit below to tip/genirq - thanks Pawel.

Ingo

------------->

Quoted 1 line

From 7e6e178ab1548c8d894a77593e757acf4510b8ba Mon Sep 17 00:00:00 2001

From: Pawel MOLL <[email protected]>
Date: Mon, 1 Sep 2008 10:12:11 +0100
Subject: [PATCH] genirq: irq_chip->startup() usage in setup_irq and set_irq_chained handler

This patch clarifies usage of irq_chip->startup() callback:

1. The "if (startup) startup(); else enabled();" code in setup_irq() is unnecessary, as startup() falls back to enabled() via default callbacks, set by irq_chip_set_defaults().

2. When using set_irq_chained_handler() the startup() was never called, which is not good at all... Fixed. And again - when startup() is not defined the call will fall back to enable() than to unmask() via default callbacks.

---
 kernel/irq/chip.c   |    2 +-
 kernel/irq/manage.c |    5 +----
 2 files changed, 2 insertions(+), 5 deletions(-)
kernel/irq/chip.c +1 -1
index 964964b..240c64d 100644
@@ -587,7 +587,7 @@ __set_irq_handler(unsigned int irq, irq_flow_handler_t handle, int is_chained,
 		desc->status &= ~IRQ_DISABLED;
 		desc->status |= IRQ_NOREQUEST | IRQ_NOPROBE;
 		desc->depth = 0;
-		desc->chip->unmask(irq);
+		desc->chip->startup(irq);
 	}
 	spin_unlock_irqrestore(&desc->lock, flags);
 }
kernel/irq/manage.c +1 -4
index ae1b684..9aa3e7b 100644
@@ -397,10 +397,7 @@ int setup_irq(unsigned int irq, struct irqaction *new)
 		if (!(desc->status & IRQ_NOAUTOEN)) {
 			desc->depth = 0;
 			desc->status &= ~IRQ_DISABLED;
-			if (desc->chip->startup)
-				desc->chip->startup(irq);
-			else
-				desc->chip->enable(irq);
+			desc->chip->startup(irq);
 		} else
 			/* Undo nested disables: */
 			desc->depth = 1;