| Message ID | 20180719193823.9486-1-u.kleine-koenig@pengutronix.de |
|---|---|
| State | New |
| Headers | show |
| Series | ARM: dts: imx6: RDU2: fix irq type for mv88e6xxx switch | expand |
Hi Uwe, On Thu, Jul 19, 2018 at 4:38 PM, Uwe Kleine-König <u.kleine-koenig@pengutronix.de> wrote: > The Marvell switches report their interrupts in a level sensitive way. > When using edge sensitive detection a race condition in the interrupt > handler of the swich might result in the OS to miss all future events > which might make the switch non-functional. > > The problem is that both mv88e6xxx_g2_irq_thread_fn() and > mv88e6xxx_g1_irq_thread_work() sample the irq cause register > (MV88E6XXX_G2_INT_SRC and MV88E6XXX_G1_STS respectively) once and then > handle the observed sources. If after sampling but before all observed > irq sources are handled a new irq source gets active this is not noticed > by the handler which returns unsuspecting, but the interrupt line stays > active which prevents the edge detector to kick in. > > All device trees but imx6qdl-zii-rdu2 get this right (most of them by > not specifying an interrupt parent). So fix imx6qdl-zii-rdu2 > accordingly. > > Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> It would be nice to add a Fixes tag.
On Thu, Jul 19, 2018 at 09:38:23PM +0200, Uwe Kleine-König wrote: > The Marvell switches report their interrupts in a level sensitive way. > When using edge sensitive detection a race condition in the interrupt > handler of the swich might result in the OS to miss all future events > which might make the switch non-functional. > > The problem is that both mv88e6xxx_g2_irq_thread_fn() and > mv88e6xxx_g1_irq_thread_work() sample the irq cause register > (MV88E6XXX_G2_INT_SRC and MV88E6XXX_G1_STS respectively) once and then > handle the observed sources. If after sampling but before all observed > irq sources are handled a new irq source gets active this is not noticed > by the handler which returns unsuspecting, but the interrupt line stays > active which prevents the edge detector to kick in. > > All device trees but imx6qdl-zii-rdu2 get this right (most of them by > not specifying an interrupt parent). So fix imx6qdl-zii-rdu2 > accordingly. > > Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> Reviewed-by: Andrew Lunn <andrew@lunn.ch> Fixes: f64992d1a916 ("ARM: dts: imx6: RDU2: Add Switch interrupts") Andrew
On Thu, Jul 19, 2018 at 10:53:58PM +0200, Andrew Lunn wrote: > On Thu, Jul 19, 2018 at 09:38:23PM +0200, Uwe Kleine-König wrote: > > The Marvell switches report their interrupts in a level sensitive way. > > When using edge sensitive detection a race condition in the interrupt > > handler of the swich might result in the OS to miss all future events > > which might make the switch non-functional. > > > > The problem is that both mv88e6xxx_g2_irq_thread_fn() and > > mv88e6xxx_g1_irq_thread_work() sample the irq cause register > > (MV88E6XXX_G2_INT_SRC and MV88E6XXX_G1_STS respectively) once and then > > handle the observed sources. If after sampling but before all observed > > irq sources are handled a new irq source gets active this is not noticed > > by the handler which returns unsuspecting, but the interrupt line stays > > active which prevents the edge detector to kick in. > > > > All device trees but imx6qdl-zii-rdu2 get this right (most of them by > > not specifying an interrupt parent). So fix imx6qdl-zii-rdu2 > > accordingly. > > > > Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> > > Reviewed-by: Andrew Lunn <andrew@lunn.ch> > Fixes: f64992d1a916 ("ARM: dts: imx6: RDU2: Add Switch interrupts") Applied with the Fixes tag added, thanks.
diff --git a/arch/arm/boot/dts/imx6qdl-zii-rdu2.dtsi b/arch/arm/boot/dts/imx6qdl-zii-rdu2.dtsi index 19a075aee19e..f14df0baf2ab 100644 --- a/arch/arm/boot/dts/imx6qdl-zii-rdu2.dtsi +++ b/arch/arm/boot/dts/imx6qdl-zii-rdu2.dtsi @@ -692,7 +692,7 @@ dsa,member = <0 0>; eeprom-length = <512>; interrupt-parent = <&gpio6>; - interrupts = <3 IRQ_TYPE_EDGE_FALLING>; + interrupts = <3 IRQ_TYPE_LEVEL_LOW>; interrupt-controller; #interrupt-cells = <2>;
The Marvell switches report their interrupts in a level sensitive way. When using edge sensitive detection a race condition in the interrupt handler of the swich might result in the OS to miss all future events which might make the switch non-functional. The problem is that both mv88e6xxx_g2_irq_thread_fn() and mv88e6xxx_g1_irq_thread_work() sample the irq cause register (MV88E6XXX_G2_INT_SRC and MV88E6XXX_G1_STS respectively) once and then handle the observed sources. If after sampling but before all observed irq sources are handled a new irq source gets active this is not noticed by the handler which returns unsuspecting, but the interrupt line stays active which prevents the edge detector to kick in. All device trees but imx6qdl-zii-rdu2 get this right (most of them by not specifying an interrupt parent). So fix imx6qdl-zii-rdu2 accordingly. Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de> --- Cc: Shawn Guo <shawnguo@kernel.org> Cc: Andrey Smirnov <andrew.smirnov@gmail.com> Cc: kernel@pengutronix.de Cc: Chris Healy <cphealy@gmail.com> Cc: Andrew Lunn <andrew@lunn.ch> Cc: Vivien Didelot <vivien.didelot@savoirfairelinux.com> Cc: Russell King <linux@armlinux.org.uk> Cc: Thomas Gleixner <tglx@linutronix.de> --- arch/arm/boot/dts/imx6qdl-zii-rdu2.dtsi | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)