[4/8] ARM: dts: imx7s: Add node for GPC

Commit Message

Tyler Baker April 13, 2017, 9:35 p.m. UTC

On 13 April 2017 at 13:49, Fabio Estevam <[e-mail]> wrote:
> On Thu, Apr 13, 2017 at 5:13 PM, Tyler Baker <[e-mail]> wrote:
>
>> I've rebased this series on the next-20170413 for sanity sake, and
>> realized there doesn't appear to be a way to select CONFIG_IMX_GPCV2.
>> I forced it using 'default y' and configured with imx_v6_v7_defconfig.
>> Now my board is booting. Before this series is applied, it may be good
>> to have CONFIG_IMX_GPCV2 selected specifically for iMX7 platform,
>> otherwise there will be boot regressions.
>
> Something like this?
>
> diff --git a/drivers/irqchip/Kconfig b/drivers/irqchip/Kconfig
> index f03ef43..fe99a48 100644
> --- a/drivers/irqchip/Kconfig
> +++ b/drivers/irqchip/Kconfig
> @@ -253,6 +253,7 @@ config RENESAS_H8S_INTC
>
>  config IMX_GPCV2
>   bool
> + def_bool y if SOC_IMX7D
>   select IRQ_DOMAIN
>   help
>    Enables the wakeup IRQs for IMX platforms with GPCv2 block

Exactly. LGTM.

>> I'd encounter a backtrace with next-20170413 +  imx_v6_v7_defconfig +
>> CONFIG_IMX_GPCV2=y
>>
>> Backtrace:
>> [<c010c364>] (dump_backtrace) from [<c010c610>] (show_stack+0x18/0x1c)
>
> but prior to fixing the Kconfig we need to fix this backtrace you reported.

I dug into this a bit, and lockdep is unhappy about spin locks not
being initialized before use.

The following patch fixes the backtrace on my board. I'll submit this
patch in a moment.

From: Tyler Baker <[e-mail]>
Date: Thu, 13 Apr 2017 14:29:49 -0700
Subject: [PATCH] irqchip/irq-imx-gpcv2: fix spinlock initialization

Call raw_spin_lock_init() before the spinlocks are used to prevent a
lockdep splat.

Signed-off-by: Tyler Baker <[e-mail]>
---
 drivers/irqchip/irq-imx-gpcv2.c | 2 ++
 1 file changed, 2 insertions(+)

  pr_err("fsl-gpcv2: unable to map gpc registers\n");

Comments

Fabio Estevam April 13, 2017, 10 p.m. UTC | #1

On Thu, Apr 13, 2017 at 6:35 PM, Tyler Baker <[e-mail]> wrote:

> The following patch fixes the backtrace on my board. I'll submit this
> patch in a moment.
>
> From: Tyler Baker <[e-mail]>
> Date: Thu, 13 Apr 2017 14:29:49 -0700
> Subject: [PATCH] irqchip/irq-imx-gpcv2: fix spinlock initialization
>
> Call raw_spin_lock_init() before the spinlocks are used to prevent a
> lockdep splat.
>
> Signed-off-by: Tyler Baker <[e-mail]>

Yes, it makes sense:

Reviewed-by: Fabio Estevam <[e-mail]>

Patch

diff --git a/drivers/irqchip/irq-imx-gpcv2.c b/drivers/irqchip/irq-imx-gpcv2.c
index e13236f..9463f35 100644
--- a/drivers/irqchip/irq-imx-gpcv2.c
+++ b/drivers/irqchip/irq-imx-gpcv2.c
@@ -230,6 +230,8 @@  static int __init imx_gpcv2_irqchip_init(struct
device_node *node,
  return -ENOMEM;
  }

+ raw_spin_lock_init(&cd->rlock);
+
  cd->gpc_base = of_iomap(node, 0);
  if (!cd->gpc_base) {