[PATCH] irqchip/atmel-aic: Fix potential deadlock in ->xlate()

[PATCH] irqchip/atmel-aic: Fix potential deadlock in ->xlate()

[Boris Brezillon]

aic5_irq_domain_xlate() and aic_irq_domain_xlate() take the generic chip
lock without disabling interrupts, which can lead to a deadlock if an
interrupt occurs while the lock is held in one of these functions.

Replace irq_gc_{lock,unlock}() calls by
raw_spin_{lock_irqsave,unlock_irqrestore}() ones to prevent this bug from
happening.

Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
Fixes: b1479ebb7720 ("irqchip: atmel-aic: Add atmel AIC/AIC5 drivers")
Cc: <stable@kernel.vger.org>
---
 drivers/irqchip/irq-atmel-aic.c  | 5 +++--
 drivers/irqchip/irq-atmel-aic5.c | 5 +++--
 2 files changed, 6 insertions(+), 4 deletions(-)

diff --git a/drivers/irqchip/irq-atmel-aic.c b/drivers/irqchip/irq-atmel-aic.c
index 112e17c2768b..447a8b0f1395 100644
--- a/drivers/irqchip/irq-atmel-aic.c
+++ b/drivers/irqchip/irq-atmel-aic.c
@@ -176,6 +176,7 @@ static int aic_irq_domain_xlate(struct irq_domain *d,
 {
 	struct irq_domain_chip_generic *dgc = d->gc;
 	struct irq_chip_generic *gc;
+	unsigned long flags;
 	unsigned smr;
 	int idx;
 	int ret;
@@ -194,11 +195,11 @@ static int aic_irq_domain_xlate(struct irq_domain *d,
 
 	gc = dgc->gc[idx];
 
-	irq_gc_lock(gc);
+	raw_spin_lock_irqsave(&gc->lock, flags);
 	smr = irq_reg_readl(gc, AT91_AIC_SMR(*out_hwirq));
 	aic_common_set_priority(intspec[2], &smr);
 	irq_reg_writel(gc, smr, AT91_AIC_SMR(*out_hwirq));
-	irq_gc_unlock(gc);
+	raw_spin_unlock_irqrestore(&gc->lock, flags);
 
 	return ret;
 }
diff --git a/drivers/irqchip/irq-atmel-aic5.c b/drivers/irqchip/irq-atmel-aic5.c
index 4f0d068e1abe..8c024107fd03 100644
--- a/drivers/irqchip/irq-atmel-aic5.c
+++ b/drivers/irqchip/irq-atmel-aic5.c
@@ -258,6 +258,7 @@ static int aic5_irq_domain_xlate(struct irq_domain *d,
 				 unsigned int *out_type)
 {
 	struct irq_chip_generic *bgc = irq_get_domain_generic_chip(d, 0);
+	unsigned long flags;
 	unsigned smr;
 	int ret;
 
@@ -269,12 +270,12 @@ static int aic5_irq_domain_xlate(struct irq_domain *d,
 	if (ret)
 		return ret;
 
-	irq_gc_lock(bgc);
+	raw_spin_lock_irqsave(&bgc->lock, flags);
 	irq_reg_writel(bgc, *out_hwirq, AT91_AIC5_SSR);
 	smr = irq_reg_readl(bgc, AT91_AIC5_SMR);
 	aic_common_set_priority(intspec[2], &smr);
 	irq_reg_writel(bgc, smr, AT91_AIC5_SMR);
-	irq_gc_unlock(bgc);
+	raw_spin_unlock_irqrestore(&bgc->lock, flags);
 
 	return ret;
 }
-- 
2.7.4

Re: [PATCH] irqchip/atmel-aic: Fix potential deadlock in ->xlate()

[Boris Brezillon]

On Tue, 13 Sep 2016 12:10:09 +0200
Boris Brezillon <boris.brezillon@free-electrons.com> wrote:

> aic5_irq_domain_xlate() and aic_irq_domain_xlate() take the generic chip
> lock without disabling interrupts, which can lead to a deadlock if an
> interrupt occurs while the lock is held in one of these functions.
> 
> Replace irq_gc_{lock,unlock}() calls by
> raw_spin_{lock_irqsave,unlock_irqrestore}() ones to prevent this bug from
> happening.
> 

Forgot to mention that this problem was discovered by kernelci [1].

[1]https://storage.kernelci.org/arm-soc/v4.4-2499-gbb46b50a5130/arm-multi_v7_defconfig+CONFIG_PROVE_LOCKING=y/lab-free-electrons/boot-at91-sama5d3_xplained.html

> Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
> Fixes: b1479ebb7720 ("irqchip: atmel-aic: Add atmel AIC/AIC5 drivers")
> Cc: <stable@kernel.vger.org>
> ---
>  drivers/irqchip/irq-atmel-aic.c  | 5 +++--
>  drivers/irqchip/irq-atmel-aic5.c | 5 +++--
>  2 files changed, 6 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/irqchip/irq-atmel-aic.c b/drivers/irqchip/irq-atmel-aic.c
> index 112e17c2768b..447a8b0f1395 100644
> --- a/drivers/irqchip/irq-atmel-aic.c
> +++ b/drivers/irqchip/irq-atmel-aic.c
> @@ -176,6 +176,7 @@ static int aic_irq_domain_xlate(struct irq_domain *d,
>  {
>  	struct irq_domain_chip_generic *dgc = d->gc;
>  	struct irq_chip_generic *gc;
> +	unsigned long flags;
>  	unsigned smr;
>  	int idx;
>  	int ret;
> @@ -194,11 +195,11 @@ static int aic_irq_domain_xlate(struct irq_domain *d,
>  
>  	gc = dgc->gc[idx];
>  
> -	irq_gc_lock(gc);
> +	raw_spin_lock_irqsave(&gc->lock, flags);
>  	smr = irq_reg_readl(gc, AT91_AIC_SMR(*out_hwirq));
>  	aic_common_set_priority(intspec[2], &smr);
>  	irq_reg_writel(gc, smr, AT91_AIC_SMR(*out_hwirq));
> -	irq_gc_unlock(gc);
> +	raw_spin_unlock_irqrestore(&gc->lock, flags);
>  
>  	return ret;
>  }
> diff --git a/drivers/irqchip/irq-atmel-aic5.c b/drivers/irqchip/irq-atmel-aic5.c
> index 4f0d068e1abe..8c024107fd03 100644
> --- a/drivers/irqchip/irq-atmel-aic5.c
> +++ b/drivers/irqchip/irq-atmel-aic5.c
> @@ -258,6 +258,7 @@ static int aic5_irq_domain_xlate(struct irq_domain *d,
>  				 unsigned int *out_type)
>  {
>  	struct irq_chip_generic *bgc = irq_get_domain_generic_chip(d, 0);
> +	unsigned long flags;
>  	unsigned smr;
>  	int ret;
>  
> @@ -269,12 +270,12 @@ static int aic5_irq_domain_xlate(struct irq_domain *d,
>  	if (ret)
>  		return ret;
>  
> -	irq_gc_lock(bgc);
> +	raw_spin_lock_irqsave(&bgc->lock, flags);
>  	irq_reg_writel(bgc, *out_hwirq, AT91_AIC5_SSR);
>  	smr = irq_reg_readl(bgc, AT91_AIC5_SMR);
>  	aic_common_set_priority(intspec[2], &smr);
>  	irq_reg_writel(bgc, smr, AT91_AIC5_SMR);
> -	irq_gc_unlock(bgc);
> +	raw_spin_unlock_irqrestore(&bgc->lock, flags);
>  
>  	return ret;
>  }

Re: [PATCH] irqchip/atmel-aic: Fix potential deadlock in ->xlate()

[Marc Zyngier]

On 13/09/16 11:10, Boris Brezillon wrote:
> aic5_irq_domain_xlate() and aic_irq_domain_xlate() take the generic chip
> lock without disabling interrupts, which can lead to a deadlock if an
> interrupt occurs while the lock is held in one of these functions.
> 
> Replace irq_gc_{lock,unlock}() calls by
> raw_spin_{lock_irqsave,unlock_irqrestore}() ones to prevent this bug from
> happening.
> 
> Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
> Fixes: b1479ebb7720 ("irqchip: atmel-aic: Add atmel AIC/AIC5 drivers")
> Cc: <stable@kernel.vger.org>
> ---
>  drivers/irqchip/irq-atmel-aic.c  | 5 +++--
>  drivers/irqchip/irq-atmel-aic5.c | 5 +++--
>  2 files changed, 6 insertions(+), 4 deletions(-)

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

Thomas, any chance you could queue this one as a fix for 4.8-rc7?

Thanks,

	M.
-- 
Jazz is not dead. It just smells funny...

Re: [PATCH] irqchip/atmel-aic: Fix potential deadlock in ->xlate()

[Boris Brezillon]

On Tue, 13 Sep 2016 12:10:09 +0200
Boris Brezillon <boris.brezillon@free-electrons.com> wrote:

> aic5_irq_domain_xlate() and aic_irq_domain_xlate() take the generic chip
> lock without disabling interrupts, which can lead to a deadlock if an
> interrupt occurs while the lock is held in one of these functions.
> 
> Replace irq_gc_{lock,unlock}() calls by
> raw_spin_{lock_irqsave,unlock_irqrestore}() ones to prevent this bug from
> happening.
> 
> Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
> Fixes: b1479ebb7720 ("irqchip: atmel-aic: Add atmel AIC/AIC5 drivers")
> Cc: <stable@kernel.vger.org>

Oops, should be

Cc: <stable@vger.kernel.org>

Do you want me to send a v2?

> ---
>  drivers/irqchip/irq-atmel-aic.c  | 5 +++--
>  drivers/irqchip/irq-atmel-aic5.c | 5 +++--
>  2 files changed, 6 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/irqchip/irq-atmel-aic.c b/drivers/irqchip/irq-atmel-aic.c
> index 112e17c2768b..447a8b0f1395 100644
> --- a/drivers/irqchip/irq-atmel-aic.c
> +++ b/drivers/irqchip/irq-atmel-aic.c
> @@ -176,6 +176,7 @@ static int aic_irq_domain_xlate(struct irq_domain *d,
>  {
>  	struct irq_domain_chip_generic *dgc = d->gc;
>  	struct irq_chip_generic *gc;
> +	unsigned long flags;
>  	unsigned smr;
>  	int idx;
>  	int ret;
> @@ -194,11 +195,11 @@ static int aic_irq_domain_xlate(struct irq_domain *d,
>  
>  	gc = dgc->gc[idx];
>  
> -	irq_gc_lock(gc);
> +	raw_spin_lock_irqsave(&gc->lock, flags);
>  	smr = irq_reg_readl(gc, AT91_AIC_SMR(*out_hwirq));
>  	aic_common_set_priority(intspec[2], &smr);
>  	irq_reg_writel(gc, smr, AT91_AIC_SMR(*out_hwirq));
> -	irq_gc_unlock(gc);
> +	raw_spin_unlock_irqrestore(&gc->lock, flags);
>  
>  	return ret;
>  }
> diff --git a/drivers/irqchip/irq-atmel-aic5.c b/drivers/irqchip/irq-atmel-aic5.c
> index 4f0d068e1abe..8c024107fd03 100644
> --- a/drivers/irqchip/irq-atmel-aic5.c
> +++ b/drivers/irqchip/irq-atmel-aic5.c
> @@ -258,6 +258,7 @@ static int aic5_irq_domain_xlate(struct irq_domain *d,
>  				 unsigned int *out_type)
>  {
>  	struct irq_chip_generic *bgc = irq_get_domain_generic_chip(d, 0);
> +	unsigned long flags;
>  	unsigned smr;
>  	int ret;
>  
> @@ -269,12 +270,12 @@ static int aic5_irq_domain_xlate(struct irq_domain *d,
>  	if (ret)
>  		return ret;
>  
> -	irq_gc_lock(bgc);
> +	raw_spin_lock_irqsave(&bgc->lock, flags);
>  	irq_reg_writel(bgc, *out_hwirq, AT91_AIC5_SSR);
>  	smr = irq_reg_readl(bgc, AT91_AIC5_SMR);
>  	aic_common_set_priority(intspec[2], &smr);
>  	irq_reg_writel(bgc, smr, AT91_AIC5_SMR);
> -	irq_gc_unlock(bgc);
> +	raw_spin_unlock_irqrestore(&bgc->lock, flags);
>  
>  	return ret;
>  }

Re: [PATCH] irqchip/atmel-aic: Fix potential deadlock in ->xlate()

[Marc Zyngier]

On 13/09/16 12:23, Boris Brezillon wrote:
> On Tue, 13 Sep 2016 12:10:09 +0200
> Boris Brezillon <boris.brezillon@free-electrons.com> wrote:
> 
>> aic5_irq_domain_xlate() and aic_irq_domain_xlate() take the generic chip
>> lock without disabling interrupts, which can lead to a deadlock if an
>> interrupt occurs while the lock is held in one of these functions.
>>
>> Replace irq_gc_{lock,unlock}() calls by
>> raw_spin_{lock_irqsave,unlock_irqrestore}() ones to prevent this bug from
>> happening.
>>
>> Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
>> Fixes: b1479ebb7720 ("irqchip: atmel-aic: Add atmel AIC/AIC5 drivers")
>> Cc: <stable@kernel.vger.org>
> 
> Oops, should be
> 
> Cc: <stable@vger.kernel.org>
> 
> Do you want me to send a v2?

You might as well, so that tglx can pick the patch with my ack on it.

Thanks,

	M.
-- 
Jazz is not dead. It just smells funny...

Re: [PATCH] irqchip/atmel-aic: Fix potential deadlock in ->xlate()

[Thomas Gleixner]

On Tue, 13 Sep 2016, Boris Brezillon wrote:

> aic5_irq_domain_xlate() and aic_irq_domain_xlate() take the generic chip
> lock without disabling interrupts, which can lead to a deadlock if an
> interrupt occurs while the lock is held in one of these functions.
> 
> Replace irq_gc_{lock,unlock}() calls by
> raw_spin_{lock_irqsave,unlock_irqrestore}() ones to prevent this bug from
> happening.
> 
> Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
> Fixes: b1479ebb7720 ("irqchip: atmel-aic: Add atmel AIC/AIC5 drivers")
> Cc: <stable@kernel.vger.org>
> ---
>  drivers/irqchip/irq-atmel-aic.c  | 5 +++--
>  drivers/irqchip/irq-atmel-aic5.c | 5 +++--
>  2 files changed, 6 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/irqchip/irq-atmel-aic.c b/drivers/irqchip/irq-atmel-aic.c
> index 112e17c2768b..447a8b0f1395 100644
> --- a/drivers/irqchip/irq-atmel-aic.c
> +++ b/drivers/irqchip/irq-atmel-aic.c
> @@ -176,6 +176,7 @@ static int aic_irq_domain_xlate(struct irq_domain *d,
>  {
>  	struct irq_domain_chip_generic *dgc = d->gc;
>  	struct irq_chip_generic *gc;
> +	unsigned long flags;
>  	unsigned smr;
>  	int idx;
>  	int ret;
> @@ -194,11 +195,11 @@ static int aic_irq_domain_xlate(struct irq_domain *d,
>  
>  	gc = dgc->gc[idx];
>  
> -	irq_gc_lock(gc);
> +	raw_spin_lock_irqsave(&gc->lock, flags);

Please provide a primitive, i.e. irq_gc_lock_irqsave(), for it.

Thanks,

	tglx

Re: [PATCH] irqchip/atmel-aic: Fix potential deadlock in ->xlate()

[Boris Brezillon]

On Tue, 13 Sep 2016 14:55:20 +0200 (CEST)
Thomas Gleixner <tglx@linutronix.de> wrote:

> On Tue, 13 Sep 2016, Boris Brezillon wrote:
> 
> > aic5_irq_domain_xlate() and aic_irq_domain_xlate() take the generic chip
> > lock without disabling interrupts, which can lead to a deadlock if an
> > interrupt occurs while the lock is held in one of these functions.
> > 
> > Replace irq_gc_{lock,unlock}() calls by
> > raw_spin_{lock_irqsave,unlock_irqrestore}() ones to prevent this bug from
> > happening.
> > 
> > Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
> > Fixes: b1479ebb7720 ("irqchip: atmel-aic: Add atmel AIC/AIC5 drivers")
> > Cc: <stable@kernel.vger.org>
> > ---
> >  drivers/irqchip/irq-atmel-aic.c  | 5 +++--
> >  drivers/irqchip/irq-atmel-aic5.c | 5 +++--
> >  2 files changed, 6 insertions(+), 4 deletions(-)
> > 
> > diff --git a/drivers/irqchip/irq-atmel-aic.c b/drivers/irqchip/irq-atmel-aic.c
> > index 112e17c2768b..447a8b0f1395 100644
> > --- a/drivers/irqchip/irq-atmel-aic.c
> > +++ b/drivers/irqchip/irq-atmel-aic.c
> > @@ -176,6 +176,7 @@ static int aic_irq_domain_xlate(struct irq_domain *d,
> >  {
> >  	struct irq_domain_chip_generic *dgc = d->gc;
> >  	struct irq_chip_generic *gc;
> > +	unsigned long flags;
> >  	unsigned smr;
> >  	int idx;
> >  	int ret;
> > @@ -194,11 +195,11 @@ static int aic_irq_domain_xlate(struct irq_domain *d,
> >  
> >  	gc = dgc->gc[idx];
> >  
> > -	irq_gc_lock(gc);
> > +	raw_spin_lock_irqsave(&gc->lock, flags);  
> 
> Please provide a primitive, i.e. irq_gc_lock_irqsave(), for it.

Sure, but that means flagging this patch for stable as well. Are you
okay with that?

> 
> Thanks,
> 
> 	tglx

Re: [PATCH] irqchip/atmel-aic: Fix potential deadlock in ->xlate()

[Thomas Gleixner]

On Tue, 13 Sep 2016, Boris Brezillon wrote:
> > Please provide a primitive, i.e. irq_gc_lock_irqsave(), for it.
> 
> Sure, but that means flagging this patch for stable as well. Are you
> okay with that?

Yes, it's still not huge ...