[PATCH] irqchip/atmel-aic[5]: Disable interrupts in the irq_pm_shutdown() callbacks

[PATCH] irqchip/atmel-aic[5]: Disable interrupts in the irq_pm_shutdown() callbacks

[Karl Mehltretter]

irq_gc_shutdown() calls the irq_pm_shutdown() callbacks from
syscore_shutdown() with interrupts enabled. aic_pm_shutdown() and
aic5_pm_shutdown() take the generic chip lock with raw_spin_lock(), and
irq_gc_eoi() takes the same lock from the interrupt handler. An
interrupt from the chip while the lock is held spins in irq_gc_eoi()
and hangs the reboot.

Lockdep reports this on every reboot of a SAM9X75 Curiosity board:

  WARNING: inconsistent lock state
  inconsistent {IN-HARDIRQ-W} -> {HARDIRQ-ON-W} usage.
  busybox/101 [HC0[0]:SC0[0]:HE1:SE1] takes:
  c1c2b434 (&gc->lock){?.-.}-{2:2}, at: aic5_pm_shutdown+0x34/0xd8
  {IN-HARDIRQ-W} state was registered at:
    ...
    irq_gc_eoi+0x28/0x60
    handle_fasteoi_irq+0x120/0x214
  ...
   aic5_pm_shutdown from irq_gc_shutdown+0x4c/0x58
   irq_gc_shutdown from syscore_shutdown+0x5c/0x9c
   syscore_shutdown from kernel_restart+0x34/0x78
   kernel_restart from sys_reboot+0x180/0x25c

Use the irqsave guard, as
commit 33517881ede7 ("irqchip/brcmstb-l2: Use _irqsave locking variants in non-interrupt code")
did for brcmstb-l2.

Fixes: b1479ebb7720 ("irqchip: atmel-aic: Add atmel AIC/AIC5 drivers")
Cc: stable@vger.kernel.org
Assisted-by: LLM
Signed-off-by: Karl Mehltretter <kmehltretter@gmail.com>
---

Notes:
    Tested on a SAM9X75 Curiosity with v7.2.8 plus this patch, PREEMPT_RT and
    non-RT lockdep builds (clang 22, LLVM=1): the report is gone. On SAM9X60
    and SAM9X7, lockdep only gets this far once the PIT64B timer report at
    boot is fixed, which I am sending separately. The irq-atmel-aic.c change
    is build-tested only.

 drivers/irqchip/irq-atmel-aic.c  | 2 +-
 drivers/irqchip/irq-atmel-aic5.c | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/irqchip/irq-atmel-aic.c b/drivers/irqchip/irq-atmel-aic.c
index 1dcc52760eca..4a7935e75e4d 100644
--- a/drivers/irqchip/irq-atmel-aic.c
+++ b/drivers/irqchip/irq-atmel-aic.c
@@ -123,7 +123,7 @@ static void aic_pm_shutdown(struct irq_data *d)
 {
 	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
 
-	guard(raw_spinlock)(&gc->lock);
+	guard(raw_spinlock_irqsave)(&gc->lock);
 	irq_reg_writel(gc, 0xffffffff, AT91_AIC_IDCR);
 	irq_reg_writel(gc, 0xffffffff, AT91_AIC_ICCR);
 }
diff --git a/drivers/irqchip/irq-atmel-aic5.c b/drivers/irqchip/irq-atmel-aic5.c
index 1f14b401f71d..07a317ed06ef 100644
--- a/drivers/irqchip/irq-atmel-aic5.c
+++ b/drivers/irqchip/irq-atmel-aic5.c
@@ -217,7 +217,7 @@ static void aic5_pm_shutdown(struct irq_data *d)
 	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
 	int i;
 
-	guard(raw_spinlock)(&bgc->lock);
+	guard(raw_spinlock_irqsave)(&bgc->lock);
 	for (i = 0; i < dgc->irqs_per_chip; i++) {
 		irq_reg_writel(bgc, i + gc->irq_base, AT91_AIC5_SSR);
 		irq_reg_writel(bgc, 1, AT91_AIC5_IDCR);
-- 
2.39.5 (Apple Git-154)

Re: [PATCH] irqchip/atmel-aic[5]: Disable interrupts in the irq_pm_shutdown() callbacks

[Radu Rendec]

On Sat, 2026-09-26 at 02:41 +0200, Karl Mehltretter wrote:
> irq_gc_shutdown() calls the irq_pm_shutdown() callbacks from
> syscore_shutdown() with interrupts enabled. aic_pm_shutdown() and
> aic5_pm_shutdown() take the generic chip lock with raw_spin_lock(), and
> irq_gc_eoi() takes the same lock from the interrupt handler. An
> interrupt from the chip while the lock is held spins in irq_gc_eoi()
> and hangs the reboot.
> 
> Lockdep reports this on every reboot of a SAM9X75 Curiosity board:
> 
>   WARNING: inconsistent lock state
>   inconsistent {IN-HARDIRQ-W} -> {HARDIRQ-ON-W} usage.
>   busybox/101 [HC0[0]:SC0[0]:HE1:SE1] takes:
>   c1c2b434 (&gc->lock){?.-.}-{2:2}, at: aic5_pm_shutdown+0x34/0xd8
>   {IN-HARDIRQ-W} state was registered at:
>     ...
>     irq_gc_eoi+0x28/0x60
>     handle_fasteoi_irq+0x120/0x214
>   ...
>    aic5_pm_shutdown from irq_gc_shutdown+0x4c/0x58
>    irq_gc_shutdown from syscore_shutdown+0x5c/0x9c
>    syscore_shutdown from kernel_restart+0x34/0x78
>    kernel_restart from sys_reboot+0x180/0x25c
> 
> Use the irqsave guard, as
> commit 33517881ede7 ("irqchip/brcmstb-l2: Use _irqsave locking variants in non-interrupt code")
> did for brcmstb-l2.
> 
> Fixes: b1479ebb7720 ("irqchip: atmel-aic: Add atmel AIC/AIC5 drivers")
> Cc: stable@vger.kernel.org
> Assisted-by: LLM
> Signed-off-by: Karl Mehltretter <kmehltretter@gmail.com>
> ---
> 
> Notes:
>     Tested on a SAM9X75 Curiosity with v7.2.8 plus this patch, PREEMPT_RT and
>     non-RT lockdep builds (clang 22, LLVM=1): the report is gone. On SAM9X60
>     and SAM9X7, lockdep only gets this far once the PIT64B timer report at
>     boot is fixed, which I am sending separately. The irq-atmel-aic.c change
>     is build-tested only.
> 
>  drivers/irqchip/irq-atmel-aic.c  | 2 +-
>  drivers/irqchip/irq-atmel-aic5.c | 2 +-
>  2 files changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/irqchip/irq-atmel-aic.c b/drivers/irqchip/irq-atmel-aic.c
> index 1dcc52760eca..4a7935e75e4d 100644
> --- a/drivers/irqchip/irq-atmel-aic.c
> +++ b/drivers/irqchip/irq-atmel-aic.c
> @@ -123,7 +123,7 @@ static void aic_pm_shutdown(struct irq_data *d)
>  {
>  	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
>  
> -	guard(raw_spinlock)(&gc->lock);
> +	guard(raw_spinlock_irqsave)(&gc->lock);
>  	irq_reg_writel(gc, 0xffffffff, AT91_AIC_IDCR);
>  	irq_reg_writel(gc, 0xffffffff, AT91_AIC_ICCR);
>  }
> diff --git a/drivers/irqchip/irq-atmel-aic5.c b/drivers/irqchip/irq-atmel-aic5.c
> index 1f14b401f71d..07a317ed06ef 100644
> --- a/drivers/irqchip/irq-atmel-aic5.c
> +++ b/drivers/irqchip/irq-atmel-aic5.c
> @@ -217,7 +217,7 @@ static void aic5_pm_shutdown(struct irq_data *d)
>  	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
>  	int i;
>  
> -	guard(raw_spinlock)(&bgc->lock);
> +	guard(raw_spinlock_irqsave)(&bgc->lock);
>  	for (i = 0; i < dgc->irqs_per_chip; i++) {
>  		irq_reg_writel(bgc, i + gc->irq_base, AT91_AIC5_SSR);
>  		irq_reg_writel(bgc, 1, AT91_AIC5_IDCR);

Reviewed-by: Radu Rendec <radu@rendec.net>