Skip to contentSkip to conversation

gpio: realtek-otto: fix GPIO line IRQ offset

PATCH In review Sander Vanheule 3 participants 4 messages Oct 28, 2021 – Oct 30, 2021 GPIO

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 gpio: realtek-otto: fix GPIO line IRQ offset

The irqchip uses one domain for all GPIO lines, so the line offset should be determined w.r.t. the first line of the first port, not the first line of the triggered port.

Fixes 0d82fb1127fb ("gpio: Add Realtek Otto GPIO support")
---
 drivers/gpio/gpio-realtek-otto.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpio/gpio-realtek-otto.c +1 -1
index eeeb39bc171d..bd75401b549d 100644
@@ -205,7 +205,7 @@ static void realtek_gpio_irq_handler(struct irq_desc *desc)
 		status = realtek_gpio_read_isr(ctrl, lines_done / 8);
 		port_pin_count = min(gc->ngpio - lines_done, 8U);
 		for_each_set_bit(offset, &status, port_pin_count)
-			generic_handle_domain_irq(gc->irq.domain, offset);
+			generic_handle_domain_irq(gc->irq.domain, offset + lines_done);
 	}
 
 	chained_irq_exit(irq_chip, desc);
Signature · 1 line
2.31.1
#1 lore raw
Quoted 6 lines · Sander Vanheule

The irqchip uses one domain for all GPIO lines, so the line offset should be determined w.r.t. the first line of the first port, not the first line of the triggered port.

Fixes 0d82fb1127fb ("gpio: Add Realtek Otto GPIO support")

Not sure it fixes anything (it wasn't working from day 1), but in any case the patch is good.

Quoted 10 lines
---
 drivers/gpio/gpio-realtek-otto.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpio/gpio-realtek-otto.c +0 -0
index eeeb39bc171d..bd75401b549d 100644

drivers/gpio/gpio-realtek-otto.c static void realtek_gpio_irq_handler(struct irq_desc *desc) · line 208

Quoted 5 lines
drivers/gpio/gpio-realtek-otto.c +1 -1
@@ -205,7 +205,7 @@ static void realtek_gpio_irq_handler(struct irq_desc *desc)
                status = realtek_gpio_read_isr(ctrl, lines_done / 8);
                port_pin_count = min(gc->ngpio - lines_done, 8U);
                for_each_set_bit(offset, &status, port_pin_count)
-                       generic_handle_domain_irq(gc->irq.domain, offset);
+                       generic_handle_domain_irq(gc->irq.domain, offset + lines_done);

Looking into these '/ 8', '8U' sounds to me that it may be a good idea to switch to for_each_set_clump8(). But it's out of scope here.

Quoted 6 lines
drivers/gpio/gpio-realtek-otto.c +0 -1
        }
 
        chained_irq_exit(irq_chip, desc);
--

2.31.1

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#2 lore raw
Quoted 10 lines · Andy Shevchenko
Quoted 6 lines · Sander Vanheule

The irqchip uses one domain for all GPIO lines, so the line offset should be determined w.r.t. the first line of the first port, not the first line of the triggered port.

Fixes 0d82fb1127fb ("gpio: Add Realtek Otto GPIO support")

Not sure it fixes anything (it wasn't working from day 1), but in any case the patch is good.

The original patch was tested using a GPIO line on port A, so the missing offset was irrelevant there (as it was zero). After recently acquiring a new device, I was trying to use a line on port C, which resulted in unhandled interrupts.

Quoted 6 lines
Quoted 6 lines

---
 drivers/gpio/gpio-realtek-otto.c | 2 +-

 1 file changed, 1 insertion(+), 1 deletion(-)
Quoted 14 lines

drivers/gpio/gpio-realtek-otto.c static void realtek_gpio_irq_handler(struct irq_desc *desc) · line 209

Quoted 11 lines
drivers/gpio/gpio-realtek-otto.c +1 -1
index eeeb39bc171d..bd75401b549d 100644
@@ -205,7 +205,7 @@ static void realtek_gpio_irq_handler(struct irq_desc *desc)
                 status = realtek_gpio_read_isr(ctrl, lines_done / 8);
                 port_pin_count = min(gc->ngpio - lines_done, 8U);
                 for_each_set_bit(offset, &status, port_pin_count)
-                       generic_handle_domain_irq(gc->irq.domain, offset);
+                       generic_handle_domain_irq(gc->irq.domain, offset +
 lines_done);

Looking into these '/ 8', '8U' sounds to me that it may be a good idea to switch to for_each_set_clump8(). But it's out of scope here.

For the current code, I could indeed have used for_each_set_clump8, instead of the u8 read_isr() accessor.

However, I'm also preparing support for the RTL9302 device I have recently acquired. On RTL930x and RTL931x the port order is reversed, requiring different port order handling. My plan is to provide different port_read_isr() functions for the different port orders, that return the requested u8. I'll try to get that patch set out today as well.

Best,
Sander

#3 lore raw
In reply to Sander Vanheule
Quoted 25 lines · Sander Vanheule

The irqchip uses one domain for all GPIO lines, so the line offset should be determined w.r.t. the first line of the first port, not the first line of the triggered port.

Fixes 0d82fb1127fb ("gpio: Add Realtek Otto GPIO support")
---
 drivers/gpio/gpio-realtek-otto.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpio/gpio-realtek-otto.c +1 -2
index eeeb39bc171d..bd75401b549d 100644
@@ -205,7 +205,7 @@ static void realtek_gpio_irq_handler(struct irq_desc *desc)
                status = realtek_gpio_read_isr(ctrl, lines_done / 8);
                port_pin_count = min(gc->ngpio - lines_done, 8U);
                for_each_set_bit(offset, &status, port_pin_count)
-                       generic_handle_domain_irq(gc->irq.domain, offset);
+                       generic_handle_domain_irq(gc->irq.domain, offset + lines_done);
        }
 
        chained_irq_exit(irq_chip, desc);
--
Quoted 2 lines

2.31.1

I already sent my last fixes PR to Linus yesterday, so I queued it with other patches for next. It'll get backported anyway to stable branches.

Bart