Skip to contentSkip to conversation

genirq: warn on inconstent flags and flow handler

PATCH Superseded Florian Fainelli 1 participant 2 messages Jun 13, 2014 – Jul 1, 2014 IRQ

v1 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 genirq: warn on inconstent flags and flow handler

It is currently possible for a generic irq chip driver to set IRQ_LEVEL and have its irq flow handler be handle_edge_irq. Setting IRQ_LEVEL in such a case does not make sense, and will actually prevent e.g: the software resend logic from kicking, and potential other problems too.

---
Thomas,

This is the change I mentioned to you on the #mipslinux IRC channel, let me know if you feel like this should be somewhere else, AFAICT this should not slow things down.

Thanks!

 kernel/irq/chip.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
kernel/irq/chip.c +6 -1
index a2b28a2fd7b1..939ae1283ec9 100644
@@ -749,8 +749,13 @@ void irq_modify_status(unsigned int irq, unsigned long clr, unsigned long set)
 		irqd_set(&desc->irq_data, IRQD_PER_CPU);
 	if (irq_settings_can_move_pcntxt(desc))
 		irqd_set(&desc->irq_data, IRQD_MOVE_PCNTXT);
-	if (irq_settings_is_level(desc))
+	if (irq_settings_is_level(desc)) {
+		/* Setting IRQD_LEVEL does not make sense on non-level
+		 * sensitive interrupts
+		 */
+		WARN_ON(desc->handle_irq != handle_level_irq);
 		irqd_set(&desc->irq_data, IRQD_LEVEL);
+	}
 
 	irqd_set(&desc->irq_data, irq_settings_get_trigger_mask(desc));
 
Signature · 1 line
1.9.1
#1 lore raw

2014-06-13 9:36 GMT-07:00 Florian Fainelli <[email protected]>:

Quoted 5 lines

It is currently possible for a generic irq chip driver to set IRQ_LEVEL and have its irq flow handler be handle_edge_irq. Setting IRQ_LEVEL in such a case does not make sense, and will actually prevent e.g: the software resend logic from kicking, and potential other problems too.

Quoted 7 lines

---
Thomas,

This is the change I mentioned to you on the #mipslinux IRC channel, let me know if you feel like this should be somewhere else, AFAICT this should not slow things down.

Thomas, Jason, is this something you are willing to accept? irq_modify_status() should not be called in a hot-path as far as I audited the code, but let me know if you prefer another solution (like propagating an error code from irq_modify_status()).

Thanks!

Quoted 28 lines

Thanks!

 kernel/irq/chip.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
kernel/irq/chip.c +6 -2
index a2b28a2fd7b1..939ae1283ec9 100644
@@ -749,8 +749,13 @@ void irq_modify_status(unsigned int irq, unsigned long clr, unsigned long set)
                irqd_set(&desc->irq_data, IRQD_PER_CPU);
        if (irq_settings_can_move_pcntxt(desc))
                irqd_set(&desc->irq_data, IRQD_MOVE_PCNTXT);
-       if (irq_settings_is_level(desc))
+       if (irq_settings_is_level(desc)) {
+               /* Setting IRQD_LEVEL does not make sense on non-level
+                * sensitive interrupts
+                */
+               WARN_ON(desc->handle_irq != handle_level_irq);
                irqd_set(&desc->irq_data, IRQD_LEVEL);
+       }
 
        irqd_set(&desc->irq_data, irq_settings_get_trigger_mask(desc));
 
--

1.9.1

Signature · 1 line
Florian