Re: [PATCH] irqchip/irq-imx-gpcv2: fix spinlock initialization

[Marc Zyngier]

On Thu, Apr 13 2017 at 11:27:31 pm BST, Tyler Baker <tyler.baker@xxxxxxxxxx> wrote:
> Call raw_spin_lock_init() before the spinlocks are used to prevent a
> lockdep splat.
>
> Fixes the following trace:
>
> INFO: trying to register non-static key.
> the code is fine but needs lockdep annotation.
> turning off the locking correctness validator.
> Hardware name: Freescale i.MX7 Dual (Device Tree)
> Backtrace:
> [<c010c364>] (dump_backtrace) from [<c010c610>] (show_stack+0x18/0x1c)
>   r7:00000000 r6:600000d3 r5:00000000 r4:c0e273dc
> [<c010c5f8>] (show_stack) from [<c04070a4>] (dump_stack+0xb4/0xe8)
> [<c0406ff0>] (dump_stack) from [<c0169e04>] (register_lock_class+0x208/0x5ec)
>  r9:ef00d010 r8:ef00d010 r7:c1606448 r6:00000000 r5:00000000 r4:ffffe000
> [<c0169bfc>] (register_lock_class) from [<c016da48>]
> (__lock_acquire+0x7c/0x18d0)
>  r10:c0e0af40 r9:ef00d010 r8:c0e274cc r7:00000001 r6:600000d3 r5:c1606448
>  r4:ffffe000
> [<c016d9cc>] (__lock_acquire) from [<c016fa4c>] (lock_acquire+0x70/0x90)
>  r10:00000000 r9:ef007e38 r8:00000001 r7:00000001 r6:600000d3 r5:00000000
>  r4:ffffe000
> [<c016f9dc>] (lock_acquire) from [<c09accc8>] (_raw_spin_lock+0x30/0x40)
>  r8:600000d3 r7:ef007e10 r6:00000001 r5:ef007e10 r4:ef00d000
> [<c09acc98>] (_raw_spin_lock) from [<c04403a4>] (imx_gpcv2_irq_unmask+0x1c/0x5c)
>  r4:ef00d000
> [<c0440388>] (imx_gpcv2_irq_unmask) from [<c017e838>] (irq_enable+0x38/0x4c)
>  r5:00000000 r4:ef007e00
> [<c017e800>] (irq_enable) from [<c017e8d0>] (irq_startup+0x84/0x88)
>  r5:00000000 r4:ef007e00
> [<c017e84c>] (irq_startup) from [<c017cd7c>] (__setup_irq+0x538/0x5f4)
>  r7:ef007e60 r6:00000015 r5:ef007e00 r4:ef007d00
> [<c017c844>] (__setup_irq) from [<c017ce98>] (setup_irq+0x60/0xd0)
>  r10:c0d5fa48 r9:efffcbc0 r8:ef007d00 r7:00000015 r6:ef007e10 r5:00000000
>  r4:ef007e00
> [<c017ce38>] (setup_irq) from [<c0d4dfdc>] (_mxc_timer_init+0x1f8/0x248)
>  r9:efffcbc0 r8:00000003 r7:016e3600 r6:c0c69bbc r5:ef007c40 r4:ef007c00
> [<c0d4dde4>] (_mxc_timer_init) from [<c0d4e0dc>] (mxc_timer_init_dt+0xb0/0xf8)
>  r7:00000000 r6:c1669e48 r5:ef7ebf7c r4:ef007c00
> [<c0d4e02c>] (mxc_timer_init_dt) from [<c0d4e168>]
> (imx6dl_timer_init_dt+0x14/0x18)
>  r9:efffcbc0 r8:c0e7b000 r7:c0c695c0 r6:c0d6fe18 r5:00000001 r4:ef7ebf7c
> [<c0d4e154>] (imx6dl_timer_init_dt) from [<c0d4d158>]
> (clocksource_probe+0x54/0xb0)
> [<c0d4d104>] (clocksource_probe) from [<c0d04a2c>] (time_init+0x30/0x38)
>  r7:c0e07900 r6:c0e7b000 r5:ffffffff r4:00000000
> [<c0d049fc>] (time_init) from [<c0d00bc8>] (start_kernel+0x220/0x3a0)
> [<c0d009a8>] (start_kernel) from [<8000807c>] (0x8000807c)
>  r10:00000000 r9:410fc075 r8:8000406a r7:c0e0c958 r6:c0d5fa44 r5:c0e07918
>  r4:c0e7b294
>
> Verified the fix on a imx7d-cl-som with CONFIG_IMX_GPCV2 set.
>
> Signed-off-by: Tyler Baker <tyler.baker@xxxxxxxxxx>
> Reported-by: Tyler Baker <tyler.baker@xxxxxxxxxx>
> Reviewed-by: Fabio Estevam <fabio.estevam@xxxxxxx>
> ---
> Issue reported in this thread: https://lkml.org/lkml/2017/4/13/646
>
>  drivers/irqchip/irq-imx-gpcv2.c | 2 ++
>  1 file changed, 2 insertions(+)
>
> 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) {
>  		pr_err("fsl-gpcv2: unable to map gpc registers\n");

Acked-by: Marc Zyngier <marc.zyngier@xxxxxxx>

Thomas, any chance you could take this as a fix through the tip tree?

Thanks,

	M.
-- 
Jazz is not dead, it just smell funny.