[BISECTED REGRESSION] v4.8-rc: gpio-leds broken on OCTEON

Hi,

gpio-leds fails to probe on OCTEON with v4.8-rc3 and when using
arch/mips/boot/dts/cavium-octeon/dlink_dsr-1000n.dts. Leds still
worked with v4.7.

I bisected this to:

	commit 15cc2ed6dcf91a8658e084be4e140147161819d7
	Author: Jon Hunter [off-list ref]
	Date:   Mon Jun 20 14:49:18 2016 +0100

	of/irq: Mark initialised interrupt controllers as populated

I have no idea how this is related to gpio-leds, except that on OCTEON
DTBs the gpio node is also interrupt controller...

Any ideas?

A.

On 08/23/2016 03:36 PM, Aaro Koskinen wrote:

Hi,

gpio-leds fails to probe on OCTEON with v4.8-rc3 and when using
arch/mips/boot/dts/cavium-octeon/dlink_dsr-1000n.dts. Leds still
worked with v4.7.

I bisected this to:

	commit 15cc2ed6dcf91a8658e084be4e140147161819d7
	Author: Jon Hunter [off-list ref]
	Date:   Mon Jun 20 14:49:18 2016 +0100

	of/irq: Mark initialised interrupt controllers as populated

I have no idea how this is related to gpio-leds, except that on OCTEON
DTBs the gpio node is also interrupt controller...

Hey Aaro.

It is actually two patches that cause the breakage. The other is:

   commit e55aeb6ba4e8cc3549bff1e75ea1d029324bce21
   of/irq: Mark interrupt controllers as populated before initialisation

I needed to revert both of these in order to get MMC working on our 71xx and 78xx boards. For our MMC, I got error messages from the MMC core of "Invalid POWER GPIO" until I applied the second patch. I will have a fix worthy of upstreaming today which will be posted today.

Steve

Hi,

On Wed, Aug 24, 2016 at 11:42:00AM -0500, Steven J. Hill wrote:

It is actually two patches that cause the breakage. The other is:

   commit e55aeb6ba4e8cc3549bff1e75ea1d029324bce21
   of/irq: Mark interrupt controllers as populated before initialisation

I needed to revert both of these in order to get MMC working on our 71xx and
78xx boards. For our MMC, I got error messages from the MMC core of "Invalid
POWER GPIO" until I applied the second patch. I will have a fix worthy of
upstreaming today which will be posted today.

The below change works for me...

diff --git a/arch/mips/cavium-octeon/octeon-irq.c b/arch/mips/cavium-octeon/octeon-irq.c
index 5a9b87b..5fd57c2 100644
--- a/arch/mips/cavium-octeon/octeon-irq.c
+++ b/arch/mips/cavium-octeon/octeon-irq.c

@@ -1618,6 +1618,7 @@ static int __init octeon_irq_init_gpio(
 		pr_warn("Cannot allocate memory for GPIO irq_domain.\n");
 		return -ENOMEM;
 	}
+	of_node_clear_flag(gpio_node, OF_POPULATED);
 
 	return 0;
 }

A.

Hi! On 25.08.2016 20:24, Aaro Koskinen wrote:

This afaics wasn't merged and the discussion looks stalled. Was this
issue discussed elsewhere or even fixed in between? Just asking, because
this issue is on the list of regressions for 4.8.

Ciao, Thorsten

Hi,

On Sun, Sep 11, 2016 at 02:41:39PM +0200, Thorsten Leemhuis wrote:

Hi! On 25.08.2016 20:24, Aaro Koskinen wrote:

quoted

On Wed, Aug 24, 2016 at 11:42:00AM -0500, Steven J. Hill wrote:

quoted

It is actually two patches that cause the breakage. The other is:
   commit e55aeb6ba4e8cc3549bff1e75ea1d029324bce21
   of/irq: Mark interrupt controllers as populated before initialisation
I needed to revert both of these in order to get MMC working on our 71xx and
78xx boards. For our MMC, I got error messages from the MMC core of "Invalid
POWER GPIO" until I applied the second patch. I will have a fix worthy of
upstreaming today which will be posted today.

The below change works for me...

diff --git a/arch/mips/cavium-octeon/octeon-irq.c b/arch/mips/cavium-octeon/octeon-irq.c
index 5a9b87b..5fd57c2 100644
--- a/arch/mips/cavium-octeon/octeon-irq.c
+++ b/arch/mips/cavium-octeon/octeon-irq.c

@@ -1618,6 +1618,7 @@ static int __init octeon_irq_init_gpio(
 		pr_warn("Cannot allocate memory for GPIO irq_domain.\n");
 		return -ENOMEM;
 	}
+	of_node_clear_flag(gpio_node, OF_POPULATED);
 
 	return 0;
 }

This afaics wasn't merged and the discussion looks stalled. Was this
issue discussed elsewhere or even fixed in between? Just asking, because
this issue is on the list of regressions for 4.8.

There's a patch waiting to be merged in Linux MIPS patchwork:

	https://patchwork.linux-mips.org/patch/14091/

A.