[PATCH] irqchip/gicv3-its: Disable the ITS before initializing it.

[PATCH] irqchip/gicv3-its: Disable the ITS before initializing it.

[David Daney]

From: David Daney <david.daney@cavium.com>

When starting a kexec/kdump kernel, the GIC ITS will already have been
enabled.  According to the ARM Generic Interrupt Controller
Architecture Specification (GIC architecture Version 3.0 and version
4.0), writing to GITS_BASER<n> or GITS_CBASER is "UNPREDICTABLE" when
the ITS is enabled.  On Cavium Thunder systems, this prevents the ITS
from being initializing in the kexec/kdump kernel, resulting in
failure to register/enable interrupts for all devices.

The fix is to disable the ITS if it is not already in the disabled
state.  This allows the ITS to be properly initialized and then
re-enabled in the kexec/kdump kernel.

Signed-off-by: David Daney <david.daney@cavium.com>
---
 drivers/irqchip/irq-gic-v3-its.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/drivers/irqchip/irq-gic-v3-its.c b/drivers/irqchip/irq-gic-v3-its.c
index 7ceaba8..36b9c28 100644
--- a/drivers/irqchip/irq-gic-v3-its.c
+++ b/drivers/irqchip/irq-gic-v3-its.c
@@ -1545,7 +1545,12 @@ static int its_force_quiescent(void __iomem *base)
 	u32 val;
 
 	val = readl_relaxed(base + GITS_CTLR);
-	if (val & GITS_CTLR_QUIESCENT)
+	/*
+	 * GIC architecture specification requires the ITS to be both
+	 * disabled and quiescent for writes to GITS_BASER<n> or
+	 * GITS_CBASER to not have UNPREDICTABLE results.
+	 */
+	if ((val & GITS_CTLR_QUIESCENT) && !(val & GITS_CTLR_ENABLE))
 		return 0;
 
 	/* Disable the generation of all interrupts to this ITS */
-- 
1.8.3.1

[PATCH] irqchip/gicv3-its: Disable the ITS before initializing it.

[Marc Zyngier]

On 18/08/16 23:41, David Daney wrote:
> From: David Daney <david.daney@cavium.com>
> 
> When starting a kexec/kdump kernel, the GIC ITS will already have been
> enabled.  According to the ARM Generic Interrupt Controller
> Architecture Specification (GIC architecture Version 3.0 and version
> 4.0), writing to GITS_BASER<n> or GITS_CBASER is "UNPREDICTABLE" when
> the ITS is enabled.  On Cavium Thunder systems, this prevents the ITS
> from being initializing in the kexec/kdump kernel, resulting in
> failure to register/enable interrupts for all devices.
> 
> The fix is to disable the ITS if it is not already in the disabled
> state.  This allows the ITS to be properly initialized and then
> re-enabled in the kexec/kdump kernel.
> 
> Signed-off-by: David Daney <david.daney@cavium.com>
> ---
>  drivers/irqchip/irq-gic-v3-its.c | 7 ++++++-
>  1 file changed, 6 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/irqchip/irq-gic-v3-its.c b/drivers/irqchip/irq-gic-v3-its.c
> index 7ceaba8..36b9c28 100644
> --- a/drivers/irqchip/irq-gic-v3-its.c
> +++ b/drivers/irqchip/irq-gic-v3-its.c
> @@ -1545,7 +1545,12 @@ static int its_force_quiescent(void __iomem *base)
>  	u32 val;
>  
>  	val = readl_relaxed(base + GITS_CTLR);
> -	if (val & GITS_CTLR_QUIESCENT)
> +	/*
> +	 * GIC architecture specification requires the ITS to be both
> +	 * disabled and quiescent for writes to GITS_BASER<n> or
> +	 * GITS_CBASER to not have UNPREDICTABLE results.
> +	 */
> +	if ((val & GITS_CTLR_QUIESCENT) && !(val & GITS_CTLR_ENABLE))
>  		return 0;
>  
>  	/* Disable the generation of all interrupts to this ITS */
> 

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

I'll queue it for the next batch of fixes that I plan to send to tglx on
Monday.

Still, there is the question of how safe it is to do a kexec on a system
that has an ITS that can modify memory behind our back, since we have no
way to tell an irqchip to disable itself altogether (not a device and
all that mess).

Thanks,

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