irqchip/sun6i-r: Use a stacked irqchip driver
The R_INTC in the A31 and newer sun8i/sun50i SoCs is more similar to the
original sun4i interrupt controller than the sun7i/sun9i NMI controller.
It is used for two distinct purposes:
- To control the trigger, latch, and mask for the NMI input pin
- To provide the interrupt input for the ARISC coprocessor
As this interrupt controller is not documented, information about it
comes from vendor-provided firmware blobs and from experimentation.
Differences from the sun4i interrupt controller appear to be:
- It only has one or two registers of each kind (max 32 or 64 IRQs)
- Multiplexing logic is added to support additional inputs
- There is no FIQ-related logic
- There is no interrupt priority logic
In order to fulfill its two purposes, this hardware block combines four
types of IRQs. First, the NMI pin is routed to the "IRQ 0" input on this
chip, with a trigger type controlled by the NMI_CTRL_REG. The "IRQ 0
pending" output from this chip, if enabled, is then routed to a SPI IRQ
input on the GIC. In other words, bit 0 of IRQ_ENABLE_REG *does* affect
the NMI IRQ seen at the GIC.
The NMI is followed by a contiguous block of 15 "direct" (my name for
them) IRQ inputs that are connected in parallel to both R_INTC and the
GIC. Or in other words, these bits of IRQ_ENABLE_REG *do not* affect the
IRQs seen at the GIC.
Following the direct IRQs are the ARISC's copy of banked IRQs for shared
peripherals. These are not relevant to Linux. The remaining IRQs are
connected to a multiplexer and provide access to the first (up to) 128
SPIs from the ARISC. This range of SPIs overlaps with the direct IRQs.
Because of the 1:1 correspondence between R_INTC and GIC inputs, this is
a perfect scenario for using a stacked irqchip driver. We want to hook
into setting the NMI trigger type, but not actually handle any IRQ here.
To allow access to all multiplexed IRQs, this driver requires a new
binding where the interrupt number matches the GIC interrupt number.
(This moves the NMI from number 0 to 32 or 96, depending on the SoC.)
For simplicity, copy the three-cell GIC binding; this disambiguates
interrupt 0 in the old binding (the NMI) from interrupt 0 in the new
binding (SPI 0) by the number of cells.
Since R_INTC is in the always-on power domain, and its output is visible
to the power management coprocessor, a stacked irqchip driver provides a
simple way to add wakeup support to any of its IRQs. That is the next
patch; for now, just the NMI is moved over.
This commit mostly reverts commit 173bda53b340 ("irqchip/sunxi-nmi:
Support sun6i-a31-r-intc compatible").
Acked-by: Maxime Ripard <mripard@kernel.org>
Signed-off-by: Samuel Holland <samuel@sholland.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20210118055040.21910-4-samuel@sholland.org
2021-01-18 08:50:33 +03:00
|
|
|
// SPDX-License-Identifier: GPL-2.0-only
|
|
|
|
/*
|
|
|
|
* The R_INTC in Allwinner A31 and newer SoCs manages several types of
|
|
|
|
* interrupts, as shown below:
|
|
|
|
*
|
|
|
|
* NMI IRQ DIRECT IRQs MUXED IRQs
|
|
|
|
* bit 0 bits 1-15^ bits 19-31
|
|
|
|
*
|
|
|
|
* +---------+ +---------+ +---------+ +---------+
|
|
|
|
* | NMI Pad | | IRQ d | | IRQ m | | IRQ m+7 |
|
|
|
|
* +---------+ +---------+ +---------+ +---------+
|
|
|
|
* | | | | | | |
|
|
|
|
* | | | | |......| |
|
|
|
|
* +------V------+ +------------+ | | | +--V------V--+ |
|
|
|
|
* | Invert/ | | Write 1 to | | | | | AND with | |
|
|
|
|
* | Edge Detect | | PENDING[0] | | | | | MUX[m/8] | |
|
|
|
|
* +-------------+ +------------+ | | | +------------+ |
|
|
|
|
* | | | | | | |
|
|
|
|
* +--V-------V--+ +--V--+ | +--V--+ | +--V--+
|
|
|
|
* | Set Reset| | GIC | | | GIC | | | GIC |
|
|
|
|
* | Latch | | SPI | | | SPI |... | ...| SPI |
|
|
|
|
* +-------------+ | N+d | | | m | | | m+7 |
|
|
|
|
* | | +-----+ | +-----+ | +-----+
|
|
|
|
* | | | |
|
|
|
|
* +-------V-+ +-V----------+ +---------V--+ +--------V--------+
|
|
|
|
* | GIC SPI | | AND with | | AND with | | AND with |
|
|
|
|
* | N (=32) | | ENABLE[0] | | ENABLE[d] | | ENABLE[19+m/8] |
|
|
|
|
* +---------+ +------------+ +------------+ +-----------------+
|
|
|
|
* | | |
|
|
|
|
* +------V-----+ +------V-----+ +--------V--------+
|
|
|
|
* | Read | | Read | | Read |
|
|
|
|
* | PENDING[0] | | PENDING[d] | | PENDING[19+m/8] |
|
|
|
|
* +------------+ +------------+ +-----------------+
|
|
|
|
*
|
|
|
|
* ^ bits 16-18 are direct IRQs for peripherals with banked interrupts, such as
|
|
|
|
* the MSGBOX. These IRQs do not map to any GIC SPI.
|
|
|
|
*
|
|
|
|
* The H6 variant adds two more (banked) direct IRQs and implements the full
|
|
|
|
* set of 128 mux bits. This requires a second set of top-level registers.
|
|
|
|
*/
|
|
|
|
|
2021-01-18 08:50:34 +03:00
|
|
|
#include <linux/bitmap.h>
|
irqchip/sun6i-r: Use a stacked irqchip driver
The R_INTC in the A31 and newer sun8i/sun50i SoCs is more similar to the
original sun4i interrupt controller than the sun7i/sun9i NMI controller.
It is used for two distinct purposes:
- To control the trigger, latch, and mask for the NMI input pin
- To provide the interrupt input for the ARISC coprocessor
As this interrupt controller is not documented, information about it
comes from vendor-provided firmware blobs and from experimentation.
Differences from the sun4i interrupt controller appear to be:
- It only has one or two registers of each kind (max 32 or 64 IRQs)
- Multiplexing logic is added to support additional inputs
- There is no FIQ-related logic
- There is no interrupt priority logic
In order to fulfill its two purposes, this hardware block combines four
types of IRQs. First, the NMI pin is routed to the "IRQ 0" input on this
chip, with a trigger type controlled by the NMI_CTRL_REG. The "IRQ 0
pending" output from this chip, if enabled, is then routed to a SPI IRQ
input on the GIC. In other words, bit 0 of IRQ_ENABLE_REG *does* affect
the NMI IRQ seen at the GIC.
The NMI is followed by a contiguous block of 15 "direct" (my name for
them) IRQ inputs that are connected in parallel to both R_INTC and the
GIC. Or in other words, these bits of IRQ_ENABLE_REG *do not* affect the
IRQs seen at the GIC.
Following the direct IRQs are the ARISC's copy of banked IRQs for shared
peripherals. These are not relevant to Linux. The remaining IRQs are
connected to a multiplexer and provide access to the first (up to) 128
SPIs from the ARISC. This range of SPIs overlaps with the direct IRQs.
Because of the 1:1 correspondence between R_INTC and GIC inputs, this is
a perfect scenario for using a stacked irqchip driver. We want to hook
into setting the NMI trigger type, but not actually handle any IRQ here.
To allow access to all multiplexed IRQs, this driver requires a new
binding where the interrupt number matches the GIC interrupt number.
(This moves the NMI from number 0 to 32 or 96, depending on the SoC.)
For simplicity, copy the three-cell GIC binding; this disambiguates
interrupt 0 in the old binding (the NMI) from interrupt 0 in the new
binding (SPI 0) by the number of cells.
Since R_INTC is in the always-on power domain, and its output is visible
to the power management coprocessor, a stacked irqchip driver provides a
simple way to add wakeup support to any of its IRQs. That is the next
patch; for now, just the NMI is moved over.
This commit mostly reverts commit 173bda53b340 ("irqchip/sunxi-nmi:
Support sun6i-a31-r-intc compatible").
Acked-by: Maxime Ripard <mripard@kernel.org>
Signed-off-by: Samuel Holland <samuel@sholland.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20210118055040.21910-4-samuel@sholland.org
2021-01-18 08:50:33 +03:00
|
|
|
#include <linux/interrupt.h>
|
|
|
|
#include <linux/irq.h>
|
|
|
|
#include <linux/irqchip.h>
|
|
|
|
#include <linux/irqdomain.h>
|
|
|
|
#include <linux/of.h>
|
|
|
|
#include <linux/of_address.h>
|
|
|
|
#include <linux/of_irq.h>
|
2021-01-18 08:50:34 +03:00
|
|
|
#include <linux/syscore_ops.h>
|
irqchip/sun6i-r: Use a stacked irqchip driver
The R_INTC in the A31 and newer sun8i/sun50i SoCs is more similar to the
original sun4i interrupt controller than the sun7i/sun9i NMI controller.
It is used for two distinct purposes:
- To control the trigger, latch, and mask for the NMI input pin
- To provide the interrupt input for the ARISC coprocessor
As this interrupt controller is not documented, information about it
comes from vendor-provided firmware blobs and from experimentation.
Differences from the sun4i interrupt controller appear to be:
- It only has one or two registers of each kind (max 32 or 64 IRQs)
- Multiplexing logic is added to support additional inputs
- There is no FIQ-related logic
- There is no interrupt priority logic
In order to fulfill its two purposes, this hardware block combines four
types of IRQs. First, the NMI pin is routed to the "IRQ 0" input on this
chip, with a trigger type controlled by the NMI_CTRL_REG. The "IRQ 0
pending" output from this chip, if enabled, is then routed to a SPI IRQ
input on the GIC. In other words, bit 0 of IRQ_ENABLE_REG *does* affect
the NMI IRQ seen at the GIC.
The NMI is followed by a contiguous block of 15 "direct" (my name for
them) IRQ inputs that are connected in parallel to both R_INTC and the
GIC. Or in other words, these bits of IRQ_ENABLE_REG *do not* affect the
IRQs seen at the GIC.
Following the direct IRQs are the ARISC's copy of banked IRQs for shared
peripherals. These are not relevant to Linux. The remaining IRQs are
connected to a multiplexer and provide access to the first (up to) 128
SPIs from the ARISC. This range of SPIs overlaps with the direct IRQs.
Because of the 1:1 correspondence between R_INTC and GIC inputs, this is
a perfect scenario for using a stacked irqchip driver. We want to hook
into setting the NMI trigger type, but not actually handle any IRQ here.
To allow access to all multiplexed IRQs, this driver requires a new
binding where the interrupt number matches the GIC interrupt number.
(This moves the NMI from number 0 to 32 or 96, depending on the SoC.)
For simplicity, copy the three-cell GIC binding; this disambiguates
interrupt 0 in the old binding (the NMI) from interrupt 0 in the new
binding (SPI 0) by the number of cells.
Since R_INTC is in the always-on power domain, and its output is visible
to the power management coprocessor, a stacked irqchip driver provides a
simple way to add wakeup support to any of its IRQs. That is the next
patch; for now, just the NMI is moved over.
This commit mostly reverts commit 173bda53b340 ("irqchip/sunxi-nmi:
Support sun6i-a31-r-intc compatible").
Acked-by: Maxime Ripard <mripard@kernel.org>
Signed-off-by: Samuel Holland <samuel@sholland.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20210118055040.21910-4-samuel@sholland.org
2021-01-18 08:50:33 +03:00
|
|
|
|
|
|
|
#include <dt-bindings/interrupt-controller/arm-gic.h>
|
|
|
|
|
|
|
|
#define SUN6I_NMI_CTRL (0x0c)
|
|
|
|
#define SUN6I_IRQ_PENDING(n) (0x10 + 4 * (n))
|
|
|
|
#define SUN6I_IRQ_ENABLE(n) (0x40 + 4 * (n))
|
|
|
|
#define SUN6I_MUX_ENABLE(n) (0xc0 + 4 * (n))
|
|
|
|
|
|
|
|
#define SUN6I_NMI_SRC_TYPE_LEVEL_LOW 0
|
|
|
|
#define SUN6I_NMI_SRC_TYPE_EDGE_FALLING 1
|
|
|
|
#define SUN6I_NMI_SRC_TYPE_LEVEL_HIGH 2
|
|
|
|
#define SUN6I_NMI_SRC_TYPE_EDGE_RISING 3
|
|
|
|
|
|
|
|
#define SUN6I_NMI_BIT BIT(0)
|
|
|
|
|
|
|
|
#define SUN6I_NMI_NEEDS_ACK ((void *)1)
|
|
|
|
|
|
|
|
#define SUN6I_NR_TOP_LEVEL_IRQS 64
|
|
|
|
#define SUN6I_NR_DIRECT_IRQS 16
|
|
|
|
#define SUN6I_NR_MUX_BITS 128
|
|
|
|
|
2021-01-18 08:50:34 +03:00
|
|
|
struct sun6i_r_intc_variant {
|
|
|
|
u32 first_mux_irq;
|
|
|
|
u32 nr_mux_irqs;
|
|
|
|
u32 mux_valid[BITS_TO_U32(SUN6I_NR_MUX_BITS)];
|
|
|
|
};
|
|
|
|
|
irqchip/sun6i-r: Use a stacked irqchip driver
The R_INTC in the A31 and newer sun8i/sun50i SoCs is more similar to the
original sun4i interrupt controller than the sun7i/sun9i NMI controller.
It is used for two distinct purposes:
- To control the trigger, latch, and mask for the NMI input pin
- To provide the interrupt input for the ARISC coprocessor
As this interrupt controller is not documented, information about it
comes from vendor-provided firmware blobs and from experimentation.
Differences from the sun4i interrupt controller appear to be:
- It only has one or two registers of each kind (max 32 or 64 IRQs)
- Multiplexing logic is added to support additional inputs
- There is no FIQ-related logic
- There is no interrupt priority logic
In order to fulfill its two purposes, this hardware block combines four
types of IRQs. First, the NMI pin is routed to the "IRQ 0" input on this
chip, with a trigger type controlled by the NMI_CTRL_REG. The "IRQ 0
pending" output from this chip, if enabled, is then routed to a SPI IRQ
input on the GIC. In other words, bit 0 of IRQ_ENABLE_REG *does* affect
the NMI IRQ seen at the GIC.
The NMI is followed by a contiguous block of 15 "direct" (my name for
them) IRQ inputs that are connected in parallel to both R_INTC and the
GIC. Or in other words, these bits of IRQ_ENABLE_REG *do not* affect the
IRQs seen at the GIC.
Following the direct IRQs are the ARISC's copy of banked IRQs for shared
peripherals. These are not relevant to Linux. The remaining IRQs are
connected to a multiplexer and provide access to the first (up to) 128
SPIs from the ARISC. This range of SPIs overlaps with the direct IRQs.
Because of the 1:1 correspondence between R_INTC and GIC inputs, this is
a perfect scenario for using a stacked irqchip driver. We want to hook
into setting the NMI trigger type, but not actually handle any IRQ here.
To allow access to all multiplexed IRQs, this driver requires a new
binding where the interrupt number matches the GIC interrupt number.
(This moves the NMI from number 0 to 32 or 96, depending on the SoC.)
For simplicity, copy the three-cell GIC binding; this disambiguates
interrupt 0 in the old binding (the NMI) from interrupt 0 in the new
binding (SPI 0) by the number of cells.
Since R_INTC is in the always-on power domain, and its output is visible
to the power management coprocessor, a stacked irqchip driver provides a
simple way to add wakeup support to any of its IRQs. That is the next
patch; for now, just the NMI is moved over.
This commit mostly reverts commit 173bda53b340 ("irqchip/sunxi-nmi:
Support sun6i-a31-r-intc compatible").
Acked-by: Maxime Ripard <mripard@kernel.org>
Signed-off-by: Samuel Holland <samuel@sholland.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20210118055040.21910-4-samuel@sholland.org
2021-01-18 08:50:33 +03:00
|
|
|
static void __iomem *base;
|
|
|
|
static irq_hw_number_t nmi_hwirq;
|
2021-01-18 08:50:34 +03:00
|
|
|
static DECLARE_BITMAP(wake_irq_enabled, SUN6I_NR_TOP_LEVEL_IRQS);
|
|
|
|
static DECLARE_BITMAP(wake_mux_enabled, SUN6I_NR_MUX_BITS);
|
|
|
|
static DECLARE_BITMAP(wake_mux_valid, SUN6I_NR_MUX_BITS);
|
irqchip/sun6i-r: Use a stacked irqchip driver
The R_INTC in the A31 and newer sun8i/sun50i SoCs is more similar to the
original sun4i interrupt controller than the sun7i/sun9i NMI controller.
It is used for two distinct purposes:
- To control the trigger, latch, and mask for the NMI input pin
- To provide the interrupt input for the ARISC coprocessor
As this interrupt controller is not documented, information about it
comes from vendor-provided firmware blobs and from experimentation.
Differences from the sun4i interrupt controller appear to be:
- It only has one or two registers of each kind (max 32 or 64 IRQs)
- Multiplexing logic is added to support additional inputs
- There is no FIQ-related logic
- There is no interrupt priority logic
In order to fulfill its two purposes, this hardware block combines four
types of IRQs. First, the NMI pin is routed to the "IRQ 0" input on this
chip, with a trigger type controlled by the NMI_CTRL_REG. The "IRQ 0
pending" output from this chip, if enabled, is then routed to a SPI IRQ
input on the GIC. In other words, bit 0 of IRQ_ENABLE_REG *does* affect
the NMI IRQ seen at the GIC.
The NMI is followed by a contiguous block of 15 "direct" (my name for
them) IRQ inputs that are connected in parallel to both R_INTC and the
GIC. Or in other words, these bits of IRQ_ENABLE_REG *do not* affect the
IRQs seen at the GIC.
Following the direct IRQs are the ARISC's copy of banked IRQs for shared
peripherals. These are not relevant to Linux. The remaining IRQs are
connected to a multiplexer and provide access to the first (up to) 128
SPIs from the ARISC. This range of SPIs overlaps with the direct IRQs.
Because of the 1:1 correspondence between R_INTC and GIC inputs, this is
a perfect scenario for using a stacked irqchip driver. We want to hook
into setting the NMI trigger type, but not actually handle any IRQ here.
To allow access to all multiplexed IRQs, this driver requires a new
binding where the interrupt number matches the GIC interrupt number.
(This moves the NMI from number 0 to 32 or 96, depending on the SoC.)
For simplicity, copy the three-cell GIC binding; this disambiguates
interrupt 0 in the old binding (the NMI) from interrupt 0 in the new
binding (SPI 0) by the number of cells.
Since R_INTC is in the always-on power domain, and its output is visible
to the power management coprocessor, a stacked irqchip driver provides a
simple way to add wakeup support to any of its IRQs. That is the next
patch; for now, just the NMI is moved over.
This commit mostly reverts commit 173bda53b340 ("irqchip/sunxi-nmi:
Support sun6i-a31-r-intc compatible").
Acked-by: Maxime Ripard <mripard@kernel.org>
Signed-off-by: Samuel Holland <samuel@sholland.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20210118055040.21910-4-samuel@sholland.org
2021-01-18 08:50:33 +03:00
|
|
|
|
|
|
|
static void sun6i_r_intc_ack_nmi(void)
|
|
|
|
{
|
|
|
|
writel_relaxed(SUN6I_NMI_BIT, base + SUN6I_IRQ_PENDING(0));
|
|
|
|
}
|
|
|
|
|
|
|
|
static void sun6i_r_intc_nmi_ack(struct irq_data *data)
|
|
|
|
{
|
|
|
|
if (irqd_get_trigger_type(data) & IRQ_TYPE_EDGE_BOTH)
|
|
|
|
sun6i_r_intc_ack_nmi();
|
|
|
|
else
|
|
|
|
data->chip_data = SUN6I_NMI_NEEDS_ACK;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void sun6i_r_intc_nmi_eoi(struct irq_data *data)
|
|
|
|
{
|
|
|
|
/* For oneshot IRQs, delay the ack until the IRQ is unmasked. */
|
|
|
|
if (data->chip_data == SUN6I_NMI_NEEDS_ACK && !irqd_irq_masked(data)) {
|
|
|
|
data->chip_data = NULL;
|
|
|
|
sun6i_r_intc_ack_nmi();
|
|
|
|
}
|
|
|
|
|
|
|
|
irq_chip_eoi_parent(data);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void sun6i_r_intc_nmi_unmask(struct irq_data *data)
|
|
|
|
{
|
|
|
|
if (data->chip_data == SUN6I_NMI_NEEDS_ACK) {
|
|
|
|
data->chip_data = NULL;
|
|
|
|
sun6i_r_intc_ack_nmi();
|
|
|
|
}
|
|
|
|
|
|
|
|
irq_chip_unmask_parent(data);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int sun6i_r_intc_nmi_set_type(struct irq_data *data, unsigned int type)
|
|
|
|
{
|
|
|
|
u32 nmi_src_type;
|
|
|
|
|
|
|
|
switch (type) {
|
|
|
|
case IRQ_TYPE_EDGE_RISING:
|
|
|
|
nmi_src_type = SUN6I_NMI_SRC_TYPE_EDGE_RISING;
|
|
|
|
break;
|
|
|
|
case IRQ_TYPE_EDGE_FALLING:
|
|
|
|
nmi_src_type = SUN6I_NMI_SRC_TYPE_EDGE_FALLING;
|
|
|
|
break;
|
|
|
|
case IRQ_TYPE_LEVEL_HIGH:
|
|
|
|
nmi_src_type = SUN6I_NMI_SRC_TYPE_LEVEL_HIGH;
|
|
|
|
break;
|
|
|
|
case IRQ_TYPE_LEVEL_LOW:
|
|
|
|
nmi_src_type = SUN6I_NMI_SRC_TYPE_LEVEL_LOW;
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
writel_relaxed(nmi_src_type, base + SUN6I_NMI_CTRL);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The "External NMI" GIC input connects to a latch inside R_INTC, not
|
|
|
|
* directly to the pin. So the GIC trigger type does not depend on the
|
|
|
|
* NMI pin trigger type.
|
|
|
|
*/
|
|
|
|
return irq_chip_set_type_parent(data, IRQ_TYPE_LEVEL_HIGH);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int sun6i_r_intc_nmi_set_irqchip_state(struct irq_data *data,
|
|
|
|
enum irqchip_irq_state which,
|
|
|
|
bool state)
|
|
|
|
{
|
|
|
|
if (which == IRQCHIP_STATE_PENDING && !state)
|
|
|
|
sun6i_r_intc_ack_nmi();
|
|
|
|
|
|
|
|
return irq_chip_set_parent_state(data, which, state);
|
|
|
|
}
|
|
|
|
|
2021-01-18 08:50:34 +03:00
|
|
|
static int sun6i_r_intc_irq_set_wake(struct irq_data *data, unsigned int on)
|
|
|
|
{
|
|
|
|
unsigned long offset_from_nmi = data->hwirq - nmi_hwirq;
|
|
|
|
|
|
|
|
if (offset_from_nmi < SUN6I_NR_DIRECT_IRQS)
|
|
|
|
assign_bit(offset_from_nmi, wake_irq_enabled, on);
|
|
|
|
else if (test_bit(data->hwirq, wake_mux_valid))
|
|
|
|
assign_bit(data->hwirq, wake_mux_enabled, on);
|
|
|
|
else
|
|
|
|
/* Not wakeup capable. */
|
|
|
|
return -EPERM;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
irqchip/sun6i-r: Use a stacked irqchip driver
The R_INTC in the A31 and newer sun8i/sun50i SoCs is more similar to the
original sun4i interrupt controller than the sun7i/sun9i NMI controller.
It is used for two distinct purposes:
- To control the trigger, latch, and mask for the NMI input pin
- To provide the interrupt input for the ARISC coprocessor
As this interrupt controller is not documented, information about it
comes from vendor-provided firmware blobs and from experimentation.
Differences from the sun4i interrupt controller appear to be:
- It only has one or two registers of each kind (max 32 or 64 IRQs)
- Multiplexing logic is added to support additional inputs
- There is no FIQ-related logic
- There is no interrupt priority logic
In order to fulfill its two purposes, this hardware block combines four
types of IRQs. First, the NMI pin is routed to the "IRQ 0" input on this
chip, with a trigger type controlled by the NMI_CTRL_REG. The "IRQ 0
pending" output from this chip, if enabled, is then routed to a SPI IRQ
input on the GIC. In other words, bit 0 of IRQ_ENABLE_REG *does* affect
the NMI IRQ seen at the GIC.
The NMI is followed by a contiguous block of 15 "direct" (my name for
them) IRQ inputs that are connected in parallel to both R_INTC and the
GIC. Or in other words, these bits of IRQ_ENABLE_REG *do not* affect the
IRQs seen at the GIC.
Following the direct IRQs are the ARISC's copy of banked IRQs for shared
peripherals. These are not relevant to Linux. The remaining IRQs are
connected to a multiplexer and provide access to the first (up to) 128
SPIs from the ARISC. This range of SPIs overlaps with the direct IRQs.
Because of the 1:1 correspondence between R_INTC and GIC inputs, this is
a perfect scenario for using a stacked irqchip driver. We want to hook
into setting the NMI trigger type, but not actually handle any IRQ here.
To allow access to all multiplexed IRQs, this driver requires a new
binding where the interrupt number matches the GIC interrupt number.
(This moves the NMI from number 0 to 32 or 96, depending on the SoC.)
For simplicity, copy the three-cell GIC binding; this disambiguates
interrupt 0 in the old binding (the NMI) from interrupt 0 in the new
binding (SPI 0) by the number of cells.
Since R_INTC is in the always-on power domain, and its output is visible
to the power management coprocessor, a stacked irqchip driver provides a
simple way to add wakeup support to any of its IRQs. That is the next
patch; for now, just the NMI is moved over.
This commit mostly reverts commit 173bda53b340 ("irqchip/sunxi-nmi:
Support sun6i-a31-r-intc compatible").
Acked-by: Maxime Ripard <mripard@kernel.org>
Signed-off-by: Samuel Holland <samuel@sholland.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20210118055040.21910-4-samuel@sholland.org
2021-01-18 08:50:33 +03:00
|
|
|
static struct irq_chip sun6i_r_intc_nmi_chip = {
|
|
|
|
.name = "sun6i-r-intc",
|
|
|
|
.irq_ack = sun6i_r_intc_nmi_ack,
|
|
|
|
.irq_mask = irq_chip_mask_parent,
|
|
|
|
.irq_unmask = sun6i_r_intc_nmi_unmask,
|
|
|
|
.irq_eoi = sun6i_r_intc_nmi_eoi,
|
|
|
|
.irq_set_affinity = irq_chip_set_affinity_parent,
|
|
|
|
.irq_set_type = sun6i_r_intc_nmi_set_type,
|
|
|
|
.irq_set_irqchip_state = sun6i_r_intc_nmi_set_irqchip_state,
|
2021-01-18 08:50:34 +03:00
|
|
|
.irq_set_wake = sun6i_r_intc_irq_set_wake,
|
|
|
|
.flags = IRQCHIP_SET_TYPE_MASKED,
|
|
|
|
};
|
|
|
|
|
|
|
|
static struct irq_chip sun6i_r_intc_wakeup_chip = {
|
|
|
|
.name = "sun6i-r-intc",
|
|
|
|
.irq_mask = irq_chip_mask_parent,
|
|
|
|
.irq_unmask = irq_chip_unmask_parent,
|
|
|
|
.irq_eoi = irq_chip_eoi_parent,
|
|
|
|
.irq_set_affinity = irq_chip_set_affinity_parent,
|
|
|
|
.irq_set_type = irq_chip_set_type_parent,
|
|
|
|
.irq_set_wake = sun6i_r_intc_irq_set_wake,
|
|
|
|
.flags = IRQCHIP_SET_TYPE_MASKED,
|
irqchip/sun6i-r: Use a stacked irqchip driver
The R_INTC in the A31 and newer sun8i/sun50i SoCs is more similar to the
original sun4i interrupt controller than the sun7i/sun9i NMI controller.
It is used for two distinct purposes:
- To control the trigger, latch, and mask for the NMI input pin
- To provide the interrupt input for the ARISC coprocessor
As this interrupt controller is not documented, information about it
comes from vendor-provided firmware blobs and from experimentation.
Differences from the sun4i interrupt controller appear to be:
- It only has one or two registers of each kind (max 32 or 64 IRQs)
- Multiplexing logic is added to support additional inputs
- There is no FIQ-related logic
- There is no interrupt priority logic
In order to fulfill its two purposes, this hardware block combines four
types of IRQs. First, the NMI pin is routed to the "IRQ 0" input on this
chip, with a trigger type controlled by the NMI_CTRL_REG. The "IRQ 0
pending" output from this chip, if enabled, is then routed to a SPI IRQ
input on the GIC. In other words, bit 0 of IRQ_ENABLE_REG *does* affect
the NMI IRQ seen at the GIC.
The NMI is followed by a contiguous block of 15 "direct" (my name for
them) IRQ inputs that are connected in parallel to both R_INTC and the
GIC. Or in other words, these bits of IRQ_ENABLE_REG *do not* affect the
IRQs seen at the GIC.
Following the direct IRQs are the ARISC's copy of banked IRQs for shared
peripherals. These are not relevant to Linux. The remaining IRQs are
connected to a multiplexer and provide access to the first (up to) 128
SPIs from the ARISC. This range of SPIs overlaps with the direct IRQs.
Because of the 1:1 correspondence between R_INTC and GIC inputs, this is
a perfect scenario for using a stacked irqchip driver. We want to hook
into setting the NMI trigger type, but not actually handle any IRQ here.
To allow access to all multiplexed IRQs, this driver requires a new
binding where the interrupt number matches the GIC interrupt number.
(This moves the NMI from number 0 to 32 or 96, depending on the SoC.)
For simplicity, copy the three-cell GIC binding; this disambiguates
interrupt 0 in the old binding (the NMI) from interrupt 0 in the new
binding (SPI 0) by the number of cells.
Since R_INTC is in the always-on power domain, and its output is visible
to the power management coprocessor, a stacked irqchip driver provides a
simple way to add wakeup support to any of its IRQs. That is the next
patch; for now, just the NMI is moved over.
This commit mostly reverts commit 173bda53b340 ("irqchip/sunxi-nmi:
Support sun6i-a31-r-intc compatible").
Acked-by: Maxime Ripard <mripard@kernel.org>
Signed-off-by: Samuel Holland <samuel@sholland.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20210118055040.21910-4-samuel@sholland.org
2021-01-18 08:50:33 +03:00
|
|
|
};
|
|
|
|
|
|
|
|
static int sun6i_r_intc_domain_translate(struct irq_domain *domain,
|
|
|
|
struct irq_fwspec *fwspec,
|
|
|
|
unsigned long *hwirq,
|
|
|
|
unsigned int *type)
|
|
|
|
{
|
|
|
|
/* Accept the old two-cell binding for the NMI only. */
|
|
|
|
if (fwspec->param_count == 2 && fwspec->param[0] == 0) {
|
|
|
|
*hwirq = nmi_hwirq;
|
|
|
|
*type = fwspec->param[1] & IRQ_TYPE_SENSE_MASK;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Otherwise this binding should match the GIC SPI binding. */
|
|
|
|
if (fwspec->param_count < 3)
|
|
|
|
return -EINVAL;
|
|
|
|
if (fwspec->param[0] != GIC_SPI)
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
*hwirq = fwspec->param[1];
|
|
|
|
*type = fwspec->param[2] & IRQ_TYPE_SENSE_MASK;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int sun6i_r_intc_domain_alloc(struct irq_domain *domain,
|
|
|
|
unsigned int virq,
|
|
|
|
unsigned int nr_irqs, void *arg)
|
|
|
|
{
|
|
|
|
struct irq_fwspec *fwspec = arg;
|
|
|
|
struct irq_fwspec gic_fwspec;
|
|
|
|
unsigned long hwirq;
|
|
|
|
unsigned int type;
|
|
|
|
int i, ret;
|
|
|
|
|
|
|
|
ret = sun6i_r_intc_domain_translate(domain, fwspec, &hwirq, &type);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
if (hwirq + nr_irqs > SUN6I_NR_MUX_BITS)
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
/* Construct a GIC-compatible fwspec from this fwspec. */
|
|
|
|
gic_fwspec = (struct irq_fwspec) {
|
|
|
|
.fwnode = domain->parent->fwnode,
|
|
|
|
.param_count = 3,
|
|
|
|
.param = { GIC_SPI, hwirq, type },
|
|
|
|
};
|
|
|
|
|
|
|
|
ret = irq_domain_alloc_irqs_parent(domain, virq, nr_irqs, &gic_fwspec);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
for (i = 0; i < nr_irqs; ++i, ++hwirq, ++virq) {
|
|
|
|
if (hwirq == nmi_hwirq) {
|
|
|
|
irq_domain_set_hwirq_and_chip(domain, virq, hwirq,
|
|
|
|
&sun6i_r_intc_nmi_chip, 0);
|
|
|
|
irq_set_handler(virq, handle_fasteoi_ack_irq);
|
|
|
|
} else {
|
2021-01-18 08:50:34 +03:00
|
|
|
irq_domain_set_hwirq_and_chip(domain, virq, hwirq,
|
|
|
|
&sun6i_r_intc_wakeup_chip, 0);
|
irqchip/sun6i-r: Use a stacked irqchip driver
The R_INTC in the A31 and newer sun8i/sun50i SoCs is more similar to the
original sun4i interrupt controller than the sun7i/sun9i NMI controller.
It is used for two distinct purposes:
- To control the trigger, latch, and mask for the NMI input pin
- To provide the interrupt input for the ARISC coprocessor
As this interrupt controller is not documented, information about it
comes from vendor-provided firmware blobs and from experimentation.
Differences from the sun4i interrupt controller appear to be:
- It only has one or two registers of each kind (max 32 or 64 IRQs)
- Multiplexing logic is added to support additional inputs
- There is no FIQ-related logic
- There is no interrupt priority logic
In order to fulfill its two purposes, this hardware block combines four
types of IRQs. First, the NMI pin is routed to the "IRQ 0" input on this
chip, with a trigger type controlled by the NMI_CTRL_REG. The "IRQ 0
pending" output from this chip, if enabled, is then routed to a SPI IRQ
input on the GIC. In other words, bit 0 of IRQ_ENABLE_REG *does* affect
the NMI IRQ seen at the GIC.
The NMI is followed by a contiguous block of 15 "direct" (my name for
them) IRQ inputs that are connected in parallel to both R_INTC and the
GIC. Or in other words, these bits of IRQ_ENABLE_REG *do not* affect the
IRQs seen at the GIC.
Following the direct IRQs are the ARISC's copy of banked IRQs for shared
peripherals. These are not relevant to Linux. The remaining IRQs are
connected to a multiplexer and provide access to the first (up to) 128
SPIs from the ARISC. This range of SPIs overlaps with the direct IRQs.
Because of the 1:1 correspondence between R_INTC and GIC inputs, this is
a perfect scenario for using a stacked irqchip driver. We want to hook
into setting the NMI trigger type, but not actually handle any IRQ here.
To allow access to all multiplexed IRQs, this driver requires a new
binding where the interrupt number matches the GIC interrupt number.
(This moves the NMI from number 0 to 32 or 96, depending on the SoC.)
For simplicity, copy the three-cell GIC binding; this disambiguates
interrupt 0 in the old binding (the NMI) from interrupt 0 in the new
binding (SPI 0) by the number of cells.
Since R_INTC is in the always-on power domain, and its output is visible
to the power management coprocessor, a stacked irqchip driver provides a
simple way to add wakeup support to any of its IRQs. That is the next
patch; for now, just the NMI is moved over.
This commit mostly reverts commit 173bda53b340 ("irqchip/sunxi-nmi:
Support sun6i-a31-r-intc compatible").
Acked-by: Maxime Ripard <mripard@kernel.org>
Signed-off-by: Samuel Holland <samuel@sholland.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20210118055040.21910-4-samuel@sholland.org
2021-01-18 08:50:33 +03:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static const struct irq_domain_ops sun6i_r_intc_domain_ops = {
|
|
|
|
.translate = sun6i_r_intc_domain_translate,
|
|
|
|
.alloc = sun6i_r_intc_domain_alloc,
|
|
|
|
.free = irq_domain_free_irqs_common,
|
|
|
|
};
|
|
|
|
|
2021-01-18 08:50:34 +03:00
|
|
|
static int sun6i_r_intc_suspend(void)
|
|
|
|
{
|
|
|
|
u32 buf[BITS_TO_U32(max(SUN6I_NR_TOP_LEVEL_IRQS, SUN6I_NR_MUX_BITS))];
|
|
|
|
int i;
|
|
|
|
|
|
|
|
/* Wake IRQs are enabled during system sleep and shutdown. */
|
|
|
|
bitmap_to_arr32(buf, wake_irq_enabled, SUN6I_NR_TOP_LEVEL_IRQS);
|
|
|
|
for (i = 0; i < BITS_TO_U32(SUN6I_NR_TOP_LEVEL_IRQS); ++i)
|
|
|
|
writel_relaxed(buf[i], base + SUN6I_IRQ_ENABLE(i));
|
|
|
|
bitmap_to_arr32(buf, wake_mux_enabled, SUN6I_NR_MUX_BITS);
|
|
|
|
for (i = 0; i < BITS_TO_U32(SUN6I_NR_MUX_BITS); ++i)
|
|
|
|
writel_relaxed(buf[i], base + SUN6I_MUX_ENABLE(i));
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
irqchip/sun6i-r: Use a stacked irqchip driver
The R_INTC in the A31 and newer sun8i/sun50i SoCs is more similar to the
original sun4i interrupt controller than the sun7i/sun9i NMI controller.
It is used for two distinct purposes:
- To control the trigger, latch, and mask for the NMI input pin
- To provide the interrupt input for the ARISC coprocessor
As this interrupt controller is not documented, information about it
comes from vendor-provided firmware blobs and from experimentation.
Differences from the sun4i interrupt controller appear to be:
- It only has one or two registers of each kind (max 32 or 64 IRQs)
- Multiplexing logic is added to support additional inputs
- There is no FIQ-related logic
- There is no interrupt priority logic
In order to fulfill its two purposes, this hardware block combines four
types of IRQs. First, the NMI pin is routed to the "IRQ 0" input on this
chip, with a trigger type controlled by the NMI_CTRL_REG. The "IRQ 0
pending" output from this chip, if enabled, is then routed to a SPI IRQ
input on the GIC. In other words, bit 0 of IRQ_ENABLE_REG *does* affect
the NMI IRQ seen at the GIC.
The NMI is followed by a contiguous block of 15 "direct" (my name for
them) IRQ inputs that are connected in parallel to both R_INTC and the
GIC. Or in other words, these bits of IRQ_ENABLE_REG *do not* affect the
IRQs seen at the GIC.
Following the direct IRQs are the ARISC's copy of banked IRQs for shared
peripherals. These are not relevant to Linux. The remaining IRQs are
connected to a multiplexer and provide access to the first (up to) 128
SPIs from the ARISC. This range of SPIs overlaps with the direct IRQs.
Because of the 1:1 correspondence between R_INTC and GIC inputs, this is
a perfect scenario for using a stacked irqchip driver. We want to hook
into setting the NMI trigger type, but not actually handle any IRQ here.
To allow access to all multiplexed IRQs, this driver requires a new
binding where the interrupt number matches the GIC interrupt number.
(This moves the NMI from number 0 to 32 or 96, depending on the SoC.)
For simplicity, copy the three-cell GIC binding; this disambiguates
interrupt 0 in the old binding (the NMI) from interrupt 0 in the new
binding (SPI 0) by the number of cells.
Since R_INTC is in the always-on power domain, and its output is visible
to the power management coprocessor, a stacked irqchip driver provides a
simple way to add wakeup support to any of its IRQs. That is the next
patch; for now, just the NMI is moved over.
This commit mostly reverts commit 173bda53b340 ("irqchip/sunxi-nmi:
Support sun6i-a31-r-intc compatible").
Acked-by: Maxime Ripard <mripard@kernel.org>
Signed-off-by: Samuel Holland <samuel@sholland.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20210118055040.21910-4-samuel@sholland.org
2021-01-18 08:50:33 +03:00
|
|
|
static void sun6i_r_intc_resume(void)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
|
|
|
/* Only the NMI is relevant during normal operation. */
|
|
|
|
writel_relaxed(SUN6I_NMI_BIT, base + SUN6I_IRQ_ENABLE(0));
|
|
|
|
for (i = 1; i < BITS_TO_U32(SUN6I_NR_TOP_LEVEL_IRQS); ++i)
|
|
|
|
writel_relaxed(0, base + SUN6I_IRQ_ENABLE(i));
|
|
|
|
}
|
|
|
|
|
2021-01-18 08:50:34 +03:00
|
|
|
static void sun6i_r_intc_shutdown(void)
|
|
|
|
{
|
|
|
|
sun6i_r_intc_suspend();
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct syscore_ops sun6i_r_intc_syscore_ops = {
|
|
|
|
.suspend = sun6i_r_intc_suspend,
|
|
|
|
.resume = sun6i_r_intc_resume,
|
|
|
|
.shutdown = sun6i_r_intc_shutdown,
|
|
|
|
};
|
|
|
|
|
irqchip/sun6i-r: Use a stacked irqchip driver
The R_INTC in the A31 and newer sun8i/sun50i SoCs is more similar to the
original sun4i interrupt controller than the sun7i/sun9i NMI controller.
It is used for two distinct purposes:
- To control the trigger, latch, and mask for the NMI input pin
- To provide the interrupt input for the ARISC coprocessor
As this interrupt controller is not documented, information about it
comes from vendor-provided firmware blobs and from experimentation.
Differences from the sun4i interrupt controller appear to be:
- It only has one or two registers of each kind (max 32 or 64 IRQs)
- Multiplexing logic is added to support additional inputs
- There is no FIQ-related logic
- There is no interrupt priority logic
In order to fulfill its two purposes, this hardware block combines four
types of IRQs. First, the NMI pin is routed to the "IRQ 0" input on this
chip, with a trigger type controlled by the NMI_CTRL_REG. The "IRQ 0
pending" output from this chip, if enabled, is then routed to a SPI IRQ
input on the GIC. In other words, bit 0 of IRQ_ENABLE_REG *does* affect
the NMI IRQ seen at the GIC.
The NMI is followed by a contiguous block of 15 "direct" (my name for
them) IRQ inputs that are connected in parallel to both R_INTC and the
GIC. Or in other words, these bits of IRQ_ENABLE_REG *do not* affect the
IRQs seen at the GIC.
Following the direct IRQs are the ARISC's copy of banked IRQs for shared
peripherals. These are not relevant to Linux. The remaining IRQs are
connected to a multiplexer and provide access to the first (up to) 128
SPIs from the ARISC. This range of SPIs overlaps with the direct IRQs.
Because of the 1:1 correspondence between R_INTC and GIC inputs, this is
a perfect scenario for using a stacked irqchip driver. We want to hook
into setting the NMI trigger type, but not actually handle any IRQ here.
To allow access to all multiplexed IRQs, this driver requires a new
binding where the interrupt number matches the GIC interrupt number.
(This moves the NMI from number 0 to 32 or 96, depending on the SoC.)
For simplicity, copy the three-cell GIC binding; this disambiguates
interrupt 0 in the old binding (the NMI) from interrupt 0 in the new
binding (SPI 0) by the number of cells.
Since R_INTC is in the always-on power domain, and its output is visible
to the power management coprocessor, a stacked irqchip driver provides a
simple way to add wakeup support to any of its IRQs. That is the next
patch; for now, just the NMI is moved over.
This commit mostly reverts commit 173bda53b340 ("irqchip/sunxi-nmi:
Support sun6i-a31-r-intc compatible").
Acked-by: Maxime Ripard <mripard@kernel.org>
Signed-off-by: Samuel Holland <samuel@sholland.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20210118055040.21910-4-samuel@sholland.org
2021-01-18 08:50:33 +03:00
|
|
|
static int __init sun6i_r_intc_init(struct device_node *node,
|
2021-01-18 08:50:34 +03:00
|
|
|
struct device_node *parent,
|
|
|
|
const struct sun6i_r_intc_variant *v)
|
irqchip/sun6i-r: Use a stacked irqchip driver
The R_INTC in the A31 and newer sun8i/sun50i SoCs is more similar to the
original sun4i interrupt controller than the sun7i/sun9i NMI controller.
It is used for two distinct purposes:
- To control the trigger, latch, and mask for the NMI input pin
- To provide the interrupt input for the ARISC coprocessor
As this interrupt controller is not documented, information about it
comes from vendor-provided firmware blobs and from experimentation.
Differences from the sun4i interrupt controller appear to be:
- It only has one or two registers of each kind (max 32 or 64 IRQs)
- Multiplexing logic is added to support additional inputs
- There is no FIQ-related logic
- There is no interrupt priority logic
In order to fulfill its two purposes, this hardware block combines four
types of IRQs. First, the NMI pin is routed to the "IRQ 0" input on this
chip, with a trigger type controlled by the NMI_CTRL_REG. The "IRQ 0
pending" output from this chip, if enabled, is then routed to a SPI IRQ
input on the GIC. In other words, bit 0 of IRQ_ENABLE_REG *does* affect
the NMI IRQ seen at the GIC.
The NMI is followed by a contiguous block of 15 "direct" (my name for
them) IRQ inputs that are connected in parallel to both R_INTC and the
GIC. Or in other words, these bits of IRQ_ENABLE_REG *do not* affect the
IRQs seen at the GIC.
Following the direct IRQs are the ARISC's copy of banked IRQs for shared
peripherals. These are not relevant to Linux. The remaining IRQs are
connected to a multiplexer and provide access to the first (up to) 128
SPIs from the ARISC. This range of SPIs overlaps with the direct IRQs.
Because of the 1:1 correspondence between R_INTC and GIC inputs, this is
a perfect scenario for using a stacked irqchip driver. We want to hook
into setting the NMI trigger type, but not actually handle any IRQ here.
To allow access to all multiplexed IRQs, this driver requires a new
binding where the interrupt number matches the GIC interrupt number.
(This moves the NMI from number 0 to 32 or 96, depending on the SoC.)
For simplicity, copy the three-cell GIC binding; this disambiguates
interrupt 0 in the old binding (the NMI) from interrupt 0 in the new
binding (SPI 0) by the number of cells.
Since R_INTC is in the always-on power domain, and its output is visible
to the power management coprocessor, a stacked irqchip driver provides a
simple way to add wakeup support to any of its IRQs. That is the next
patch; for now, just the NMI is moved over.
This commit mostly reverts commit 173bda53b340 ("irqchip/sunxi-nmi:
Support sun6i-a31-r-intc compatible").
Acked-by: Maxime Ripard <mripard@kernel.org>
Signed-off-by: Samuel Holland <samuel@sholland.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20210118055040.21910-4-samuel@sholland.org
2021-01-18 08:50:33 +03:00
|
|
|
{
|
|
|
|
struct irq_domain *domain, *parent_domain;
|
|
|
|
struct of_phandle_args nmi_parent;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
/* Extract the NMI hwirq number from the OF node. */
|
|
|
|
ret = of_irq_parse_one(node, 0, &nmi_parent);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
if (nmi_parent.args_count < 3 ||
|
|
|
|
nmi_parent.args[0] != GIC_SPI ||
|
|
|
|
nmi_parent.args[2] != IRQ_TYPE_LEVEL_HIGH)
|
|
|
|
return -EINVAL;
|
|
|
|
nmi_hwirq = nmi_parent.args[1];
|
|
|
|
|
2021-01-18 08:50:34 +03:00
|
|
|
bitmap_set(wake_irq_enabled, v->first_mux_irq, v->nr_mux_irqs);
|
|
|
|
bitmap_from_arr32(wake_mux_valid, v->mux_valid, SUN6I_NR_MUX_BITS);
|
|
|
|
|
irqchip/sun6i-r: Use a stacked irqchip driver
The R_INTC in the A31 and newer sun8i/sun50i SoCs is more similar to the
original sun4i interrupt controller than the sun7i/sun9i NMI controller.
It is used for two distinct purposes:
- To control the trigger, latch, and mask for the NMI input pin
- To provide the interrupt input for the ARISC coprocessor
As this interrupt controller is not documented, information about it
comes from vendor-provided firmware blobs and from experimentation.
Differences from the sun4i interrupt controller appear to be:
- It only has one or two registers of each kind (max 32 or 64 IRQs)
- Multiplexing logic is added to support additional inputs
- There is no FIQ-related logic
- There is no interrupt priority logic
In order to fulfill its two purposes, this hardware block combines four
types of IRQs. First, the NMI pin is routed to the "IRQ 0" input on this
chip, with a trigger type controlled by the NMI_CTRL_REG. The "IRQ 0
pending" output from this chip, if enabled, is then routed to a SPI IRQ
input on the GIC. In other words, bit 0 of IRQ_ENABLE_REG *does* affect
the NMI IRQ seen at the GIC.
The NMI is followed by a contiguous block of 15 "direct" (my name for
them) IRQ inputs that are connected in parallel to both R_INTC and the
GIC. Or in other words, these bits of IRQ_ENABLE_REG *do not* affect the
IRQs seen at the GIC.
Following the direct IRQs are the ARISC's copy of banked IRQs for shared
peripherals. These are not relevant to Linux. The remaining IRQs are
connected to a multiplexer and provide access to the first (up to) 128
SPIs from the ARISC. This range of SPIs overlaps with the direct IRQs.
Because of the 1:1 correspondence between R_INTC and GIC inputs, this is
a perfect scenario for using a stacked irqchip driver. We want to hook
into setting the NMI trigger type, but not actually handle any IRQ here.
To allow access to all multiplexed IRQs, this driver requires a new
binding where the interrupt number matches the GIC interrupt number.
(This moves the NMI from number 0 to 32 or 96, depending on the SoC.)
For simplicity, copy the three-cell GIC binding; this disambiguates
interrupt 0 in the old binding (the NMI) from interrupt 0 in the new
binding (SPI 0) by the number of cells.
Since R_INTC is in the always-on power domain, and its output is visible
to the power management coprocessor, a stacked irqchip driver provides a
simple way to add wakeup support to any of its IRQs. That is the next
patch; for now, just the NMI is moved over.
This commit mostly reverts commit 173bda53b340 ("irqchip/sunxi-nmi:
Support sun6i-a31-r-intc compatible").
Acked-by: Maxime Ripard <mripard@kernel.org>
Signed-off-by: Samuel Holland <samuel@sholland.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20210118055040.21910-4-samuel@sholland.org
2021-01-18 08:50:33 +03:00
|
|
|
parent_domain = irq_find_host(parent);
|
|
|
|
if (!parent_domain) {
|
|
|
|
pr_err("%pOF: Failed to obtain parent domain\n", node);
|
|
|
|
return -ENXIO;
|
|
|
|
}
|
|
|
|
|
|
|
|
base = of_io_request_and_map(node, 0, NULL);
|
|
|
|
if (IS_ERR(base)) {
|
|
|
|
pr_err("%pOF: Failed to map MMIO region\n", node);
|
|
|
|
return PTR_ERR(base);
|
|
|
|
}
|
|
|
|
|
|
|
|
domain = irq_domain_add_hierarchy(parent_domain, 0, 0, node,
|
|
|
|
&sun6i_r_intc_domain_ops, NULL);
|
|
|
|
if (!domain) {
|
|
|
|
pr_err("%pOF: Failed to allocate domain\n", node);
|
|
|
|
iounmap(base);
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
|
2021-01-18 08:50:34 +03:00
|
|
|
register_syscore_ops(&sun6i_r_intc_syscore_ops);
|
|
|
|
|
irqchip/sun6i-r: Use a stacked irqchip driver
The R_INTC in the A31 and newer sun8i/sun50i SoCs is more similar to the
original sun4i interrupt controller than the sun7i/sun9i NMI controller.
It is used for two distinct purposes:
- To control the trigger, latch, and mask for the NMI input pin
- To provide the interrupt input for the ARISC coprocessor
As this interrupt controller is not documented, information about it
comes from vendor-provided firmware blobs and from experimentation.
Differences from the sun4i interrupt controller appear to be:
- It only has one or two registers of each kind (max 32 or 64 IRQs)
- Multiplexing logic is added to support additional inputs
- There is no FIQ-related logic
- There is no interrupt priority logic
In order to fulfill its two purposes, this hardware block combines four
types of IRQs. First, the NMI pin is routed to the "IRQ 0" input on this
chip, with a trigger type controlled by the NMI_CTRL_REG. The "IRQ 0
pending" output from this chip, if enabled, is then routed to a SPI IRQ
input on the GIC. In other words, bit 0 of IRQ_ENABLE_REG *does* affect
the NMI IRQ seen at the GIC.
The NMI is followed by a contiguous block of 15 "direct" (my name for
them) IRQ inputs that are connected in parallel to both R_INTC and the
GIC. Or in other words, these bits of IRQ_ENABLE_REG *do not* affect the
IRQs seen at the GIC.
Following the direct IRQs are the ARISC's copy of banked IRQs for shared
peripherals. These are not relevant to Linux. The remaining IRQs are
connected to a multiplexer and provide access to the first (up to) 128
SPIs from the ARISC. This range of SPIs overlaps with the direct IRQs.
Because of the 1:1 correspondence between R_INTC and GIC inputs, this is
a perfect scenario for using a stacked irqchip driver. We want to hook
into setting the NMI trigger type, but not actually handle any IRQ here.
To allow access to all multiplexed IRQs, this driver requires a new
binding where the interrupt number matches the GIC interrupt number.
(This moves the NMI from number 0 to 32 or 96, depending on the SoC.)
For simplicity, copy the three-cell GIC binding; this disambiguates
interrupt 0 in the old binding (the NMI) from interrupt 0 in the new
binding (SPI 0) by the number of cells.
Since R_INTC is in the always-on power domain, and its output is visible
to the power management coprocessor, a stacked irqchip driver provides a
simple way to add wakeup support to any of its IRQs. That is the next
patch; for now, just the NMI is moved over.
This commit mostly reverts commit 173bda53b340 ("irqchip/sunxi-nmi:
Support sun6i-a31-r-intc compatible").
Acked-by: Maxime Ripard <mripard@kernel.org>
Signed-off-by: Samuel Holland <samuel@sholland.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20210118055040.21910-4-samuel@sholland.org
2021-01-18 08:50:33 +03:00
|
|
|
sun6i_r_intc_ack_nmi();
|
|
|
|
sun6i_r_intc_resume();
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
2021-01-18 08:50:34 +03:00
|
|
|
|
|
|
|
static const struct sun6i_r_intc_variant sun6i_a31_r_intc_variant __initconst = {
|
|
|
|
.first_mux_irq = 19,
|
|
|
|
.nr_mux_irqs = 13,
|
|
|
|
.mux_valid = { 0xffffffff, 0xfff80000, 0xffffffff, 0x0000000f },
|
|
|
|
};
|
|
|
|
|
|
|
|
static int __init sun6i_a31_r_intc_init(struct device_node *node,
|
|
|
|
struct device_node *parent)
|
|
|
|
{
|
|
|
|
return sun6i_r_intc_init(node, parent, &sun6i_a31_r_intc_variant);
|
|
|
|
}
|
|
|
|
IRQCHIP_DECLARE(sun6i_a31_r_intc, "allwinner,sun6i-a31-r-intc", sun6i_a31_r_intc_init);
|
|
|
|
|
|
|
|
static const struct sun6i_r_intc_variant sun50i_h6_r_intc_variant __initconst = {
|
|
|
|
.first_mux_irq = 21,
|
|
|
|
.nr_mux_irqs = 16,
|
|
|
|
.mux_valid = { 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff },
|
|
|
|
};
|
|
|
|
|
|
|
|
static int __init sun50i_h6_r_intc_init(struct device_node *node,
|
|
|
|
struct device_node *parent)
|
|
|
|
{
|
|
|
|
return sun6i_r_intc_init(node, parent, &sun50i_h6_r_intc_variant);
|
|
|
|
}
|
|
|
|
IRQCHIP_DECLARE(sun50i_h6_r_intc, "allwinner,sun50i-h6-r-intc", sun50i_h6_r_intc_init);
|