[PATCH] staging: greybus: Fix the irq API abuse

[PATCH] staging: greybus: Fix the irq API abuse

[Thomas Gleixner]

Nothing outside of low level architecture code is supposed to look up
interrupt descriptors and fiddle with them.

Replace the open coded abuse by calling generic_handle_irq().

This still does not explain why and in which context this connection
magic is injecting interrupts in the first place and why this is correct
and safe, but at least the API abuse is gone.

Fixes: 036aad9d0224 ("greybus: gpio: add interrupt handling support")
Fixes: 2611ebef8322 ("greybus: gpio: don't call irq-flow handler directly")
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
---
 drivers/staging/greybus/gpio.c |   15 ++++++---------
 1 file changed, 6 insertions(+), 9 deletions(-)

--- a/drivers/staging/greybus/gpio.c
+++ b/drivers/staging/greybus/gpio.c
@@ -364,8 +364,7 @@ static int gb_gpio_request_handler(struc
 	struct gb_message *request;
 	struct gb_gpio_irq_event_request *event;
 	u8 type = op->type;
-	int irq;
-	struct irq_desc *desc;
+	int irq, ret;
 
 	if (type != GB_GPIO_TYPE_IRQ_EVENT) {
 		dev_err(dev, "unsupported unsolicited request: %u\n", type);
@@ -391,17 +390,15 @@ static int gb_gpio_request_handler(struc
 		dev_err(dev, "failed to find IRQ\n");
 		return -EINVAL;
 	}
-	desc = irq_to_desc(irq);
-	if (!desc) {
-		dev_err(dev, "failed to look up irq\n");
-		return -EINVAL;
-	}
 
 	local_irq_disable();
-	generic_handle_irq_desc(desc);
+	ret = generic_handle_irq(irq);
 	local_irq_enable();
 
-	return 0;
+	if (ret)
+		dev_err(dev, "failed to invoke irq handler\n");
+
+	return ret;
 }
 
 static int gb_gpio_request(struct gpio_chip *chip, unsigned int offset)

Re: [PATCH] staging: greybus: Fix the irq API abuse

[Johan Hovold]

On Fri, Mar 06, 2020 at 02:24:13PM +0100, Thomas Gleixner wrote:
> Nothing outside of low level architecture code is supposed to look up
> interrupt descriptors and fiddle with them.
> 
> Replace the open coded abuse by calling generic_handle_irq().
> 
> This still does not explain why and in which context this connection
> magic is injecting interrupts in the first place and why this is correct
> and safe, but at least the API abuse is gone.

Yeah, there's more to that story. The interrupt-handling was known to
have issues, but I can't seem to find the details right now.

> Fixes: 036aad9d0224 ("greybus: gpio: add interrupt handling support")
> Fixes: 2611ebef8322 ("greybus: gpio: don't call irq-flow handler directly")
> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

Acked-by: Johan Hovold <johan@kernel.org>

Johan