Skip to contentSkip to conversation

irqchip: add support for Marvell Orion SoCs

PATCH Superseded Sebastian Hesselbarth 16 participants 84 messages May 2, 2013 – Oct 1, 2013 1 message not in the archive

Subsystems (8)

v1 of 3 Superseded Latest is v3 →

lore
Share & tools
lore

Sebastian Hesselbarth

author maintainer
#0 lore raw

PATCH 1/1 irqchip: add support for Marvell Orion SoCs

This patch adds an irqchip driver for the main interrupt controller found on Marvell Orion SoCs (Kirkwood, Dove, Orion5x, Discovery Innovation). Corresponding device tree documentation is also added.

---
Note: This patch triggers a checkpatch warning for
WARNING: Avoid CamelCase: <handle_IRQ>

Cc Grant Likely <[email protected]>
Cc Rob Herring <[email protected]>
Cc Rob Landley <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc Russell King <[email protected]>
Cc Arnd Bergmann <[email protected]>
Cc Jason Cooper <[email protected]>
Cc Andrew Lunn <[email protected]>
Cc Thomas Petazzoni <[email protected]>
Cc Gregory Clement <[email protected]>
Cc Ezequiel Garcia <[email protected]>
Cc Jean-Francois Moine <[email protected]>
Cc devicetree-discuss at lists.ozlabs.org
Cc linux-doc at vger.kernel.org
Cc linux-arm-kernel at lists.infradead.org
Cc linux-kernel at vger.kernel.org
---
 .../interrupt-controller/marvell,orion-mpic.txt    |   22 ++++
 drivers/irqchip/Kconfig                            |    5 +
 drivers/irqchip/Makefile                           |    1 +
 drivers/irqchip/irq-orion.c                        |  129 ++++++++++++++++++++
 include/linux/irqchip/orion.h                      |   18 +++
 5 files changed, 175 insertions(+), 0 deletions(-)
 create mode 100644 Documentation/devicetree/bindings/interrupt-controller/marvell,orion-mpic.txt
 create mode 100644 drivers/irqchip/irq-orion.c
 create mode 100644 include/linux/irqchip/orion.h
Documentation/devicetree/bindings/interrupt-controller/marvell,orion-mpic.txt new file +22 -0
new file mode 100644
index 0000000..3b303ec
@@ -0,0 +1,22 @@
+Marvell Orion SoC main interrupt controller
+
+Required properties:
+- compatible: shall be "marvell,orion-mpic"
+- reg: base address(es) of interrupt registers starting with CAUSE register
+- interrupt-controller: identifies the node as an interrupt controller
+- #interrupt-cells: number of cells to encode an interrupt source, shall be 1.
+
+The interrupt sources map to the corresponding bits in the interrupt
+registers, i.e.
+- 0 maps to bit 0 of first base address,
+- 1 maps to bit 1 of first base address,
+- 32 maps to bit 0 of second base address, and so on.
+
+Example:
+	intc: interrupt-controller {
+		compatible = "marvell,orion-mpic";
+		interrupt-controller;
+		#interrupt-cells = <1>;
+                /* Dove has 64 primary interrupts */
+		reg = <0x20200 0x10>, <0x20210 0x10>;
+	};
drivers/irqchip/Kconfig +5 -0
index a350969..8da3559 100644
@@ -2,6 +2,11 @@ config IRQCHIP
 	def_bool y
 	depends on OF_IRQ
 
+config IRQCHIP_ORION
+	bool
+	select IRQ_DOMAIN
+	select MULTI_IRQ_HANDLER
+
 config ARM_GIC
 	bool
 	select IRQ_DOMAIN
drivers/irqchip/Makefile +1 -0
index 98e3b87..8adbd43 100644
@@ -2,6 +2,7 @@ obj-$(CONFIG_IRQCHIP)			+= irqchip.o
 
 obj-$(CONFIG_ARCH_BCM2835)		+= irq-bcm2835.o
 obj-$(CONFIG_ARCH_EXYNOS)		+= exynos-combiner.o
+obj-$(CONFIG_IRQCHIP_ORION)		+= irq-orion.o
 obj-$(CONFIG_METAG)			+= irq-metag-ext.o
 obj-$(CONFIG_METAG_PERFCOUNTER_IRQS)	+= irq-metag.o
 obj-$(CONFIG_ARCH_SUNXI)		+= irq-sunxi.o
drivers/irqchip/irq-orion.c new file +129 -0

129 lines not on this page · raw message

include/linux/irqchip/orion.h new file +18 -0
new file mode 100644
index 0000000..04f7bab
@@ -0,0 +1,18 @@
+/*
+ * Marvell Orion SoCs IRQ chip driver header.
+ *
+ * Sebastian Hesselbarth <[email protected]>
+ *
+ * This file is licensed under the terms of the GNU General Public
+ * License version 2.  This program is licensed "as is" without any
+ * warranty of any kind, whether express or implied.
+ */
+
+#ifndef __LINUX_IRQCHIP_ORION_H
+#define __LINUX_IRQCHIP_ORION_H
+
+#include <asm/exception.h>
+
+extern void orion_init_irq(void);
+
+#endif
Signature · 1 line
1.7.2.5
author maintainer
#1 lore raw
Quoted 9 lines · Sebastian Hesselbarth

This patch adds an irqchip driver for the main interrupt controller found on Marvell Orion SoCs (Kirkwood, Dove, Orion5x, Discovery Innovation). Corresponding device tree documentation is also added.

---
Note: This patch triggers a checkpatch warning for
WARNING: Avoid CamelCase:<handle_IRQ>

[...]

Quoted 10 lines
include/linux/irqchip/orion.h new file +4 -0
new file mode 100644
index 0000000..04f7bab
@@ -0,0 +1,18 @@
+/*
+ * Marvell Orion SoCs IRQ chip driver header.
+ *
+ * Sebastian Hesselbarth<[email protected]>

include/linux/irqchip/orion.h line 14

Quoted 10 lines
include/linux/irqchip/orion.h new file +10 -0
+ *
+ * This file is licensed under the terms of the GNU General Public
+ * License version 2.  This program is licensed "as is" without any
+ * warranty of any kind, whether express or implied.
+ */
+
+#ifndef __LINUX_IRQCHIP_ORION_H
+#define __LINUX_IRQCHIP_ORION_H
+
+#include<asm/exception.h>

First review by myself. The above include is a left-over and will be removed in a v2.

Quoted 4 lines
include/linux/irqchip/orion.h new file +4 -0
+
+extern void orion_init_irq(void);
+
+#endif
#2 lore raw
Quoted 26 lines · Sebastian Hesselbarth
Quoted 9 lines · Sebastian Hesselbarth

This patch adds an irqchip driver for the main interrupt controller found on Marvell Orion SoCs (Kirkwood, Dove, Orion5x, Discovery Innovation). Corresponding device tree documentation is also added.

---
Note: This patch triggers a checkpatch warning for
WARNING: Avoid CamelCase:<handle_IRQ>

[...]

Quoted 15 lines
include/linux/irqchip/orion.h new file +9 -0
new file mode 100644
index 0000000..04f7bab
@@ -0,0 +1,18 @@
+/*
+ * Marvell Orion SoCs IRQ chip driver header.
+ *
+ * Sebastian Hesselbarth<[email protected]>
+ *
+ * This file is licensed under the terms of the GNU General Public
+ * License version 2.  This program is licensed "as is" without any
+ * warranty of any kind, whether express or implied.
+ */
Quoted 8 lines

include/linux/irqchip/orion.h

Quoted 5 lines
include/linux/irqchip/orion.h new file +5 -0
+
+#ifndef __LINUX_IRQCHIP_ORION_H
+#define __LINUX_IRQCHIP_ORION_H
+
+#include<asm/exception.h>

First review by myself. The above include is a left-over and will be removed in a v2.

You still need your first level IRQ handlers marked with __exception_irq_entry which is defined in the above file.

author maintainer
#3 lore raw
Quoted 36 lines · Russell King - ARM Linux
Quoted 26 lines · Sebastian Hesselbarth
Quoted 9 lines · Sebastian Hesselbarth

This patch adds an irqchip driver for the main interrupt controller found on Marvell Orion SoCs (Kirkwood, Dove, Orion5x, Discovery Innovation). Corresponding device tree documentation is also added.

---
Note: This patch triggers a checkpatch warning for
WARNING: Avoid CamelCase:<handle_IRQ>

[...]

Quoted 15 lines
include/linux/irqchip/orion.h new file +9 -0
new file mode 100644
index 0000000..04f7bab
@@ -0,0 +1,18 @@
+/*
+ * Marvell Orion SoCs IRQ chip driver header.
+ *
+ * Sebastian Hesselbarth<[email protected]>
+ *
+ * This file is licensed under the terms of the GNU General Public
+ * License version 2.  This program is licensed "as is" without any
+ * warranty of any kind, whether express or implied.
+ */
Quoted 8 lines

include/linux/irqchip/orion.h

Quoted 5 lines
include/linux/irqchip/orion.h new file +5 -0
+
+#ifndef __LINUX_IRQCHIP_ORION_H
+#define __LINUX_IRQCHIP_ORION_H
+
+#include<asm/exception.h>

First review by myself. The above include is a left-over and will be removed in a v2.

Quoted 3 lines

You still need your first level IRQ handlers marked with __exception_irq_entry which is defined in the above file.

Russell,

I know and it is marked with __exception_irq_entry. The above is in include/linux/irqchip/orion.h and only used for .init_irq in machine descriptor later.

The irq handler is never exposed to the board file itself, but set within orion_init_irq. This approach has been taked by irqchip/irq-gic.c and irqchip/irq-vic.c rather than adding .handle_irq to the machine descriptor.

Sebastian

#4 lore raw
Quoted 43 lines · Sebastian Hesselbarth
Quoted 36 lines · Russell King - ARM Linux
Quoted 26 lines · Sebastian Hesselbarth
Quoted 9 lines · Sebastian Hesselbarth

This patch adds an irqchip driver for the main interrupt controller found on Marvell Orion SoCs (Kirkwood, Dove, Orion5x, Discovery Innovation). Corresponding device tree documentation is also added.

---
Note: This patch triggers a checkpatch warning for
WARNING: Avoid CamelCase:<handle_IRQ>

[...]

Quoted 15 lines
include/linux/irqchip/orion.h new file +9 -0
new file mode 100644
index 0000000..04f7bab
@@ -0,0 +1,18 @@
+/*
+ * Marvell Orion SoCs IRQ chip driver header.
+ *
+ * Sebastian Hesselbarth<[email protected]>
+ *
+ * This file is licensed under the terms of the GNU General Public
+ * License version 2.  This program is licensed "as is" without any
+ * warranty of any kind, whether express or implied.
+ */
Quoted 8 lines

include/linux/irqchip/orion.h

Quoted 5 lines
include/linux/irqchip/orion.h new file +5 -0
+
+#ifndef __LINUX_IRQCHIP_ORION_H
+#define __LINUX_IRQCHIP_ORION_H
+
+#include<asm/exception.h>

First review by myself. The above include is a left-over and will be removed in a v2.

Quoted 3 lines

You still need your first level IRQ handlers marked with __exception_irq_entry which is defined in the above file.

Russell,

Quoted 8 lines

I know and it is marked with __exception_irq_entry. The above is in include/linux/irqchip/orion.h and only used for .init_irq in machine descriptor later.

The irq handler is never exposed to the board file itself, but set within orion_init_irq. This approach has been taked by irqchip/irq-gic.c and irqchip/irq-vic.c rather than adding .handle_irq to the machine descriptor.

But I don't find an asm/exception.h include in drivers/irqchip/whateveryour.cfileiscalled

author maintainer
#5 lore raw
Quoted 48 lines · Russell King - ARM Linux
Quoted 47 lines · Sebastian Hesselbarth
Quoted 36 lines · Russell King - ARM Linux
Quoted 26 lines · Sebastian Hesselbarth
Quoted 9 lines · Sebastian Hesselbarth

This patch adds an irqchip driver for the main interrupt controller found on Marvell Orion SoCs (Kirkwood, Dove, Orion5x, Discovery Innovation). Corresponding device tree documentation is also added.

---
Note: This patch triggers a checkpatch warning for
WARNING: Avoid CamelCase:<handle_IRQ>

[...]

Quoted 15 lines
include/linux/irqchip/orion.h new file +9 -0
new file mode 100644
index 0000000..04f7bab
@@ -0,0 +1,18 @@
+/*
+ * Marvell Orion SoCs IRQ chip driver header.
+ *
+ * Sebastian Hesselbarth<[email protected]>
+ *
+ * This file is licensed under the terms of the GNU General Public
+ * License version 2.  This program is licensed "as is" without any
+ * warranty of any kind, whether express or implied.
+ */
Quoted 8 lines

include/linux/irqchip/orion.h

Quoted 5 lines
include/linux/irqchip/orion.h new file +5 -0
+
+#ifndef __LINUX_IRQCHIP_ORION_H
+#define __LINUX_IRQCHIP_ORION_H
+
+#include<asm/exception.h>

First review by myself. The above include is a left-over and will be removed in a v2.

Quoted 3 lines

You still need your first level IRQ handlers marked with __exception_irq_entry which is defined in the above file.

Russell,

I know and it is marked with __exception_irq_entry. The above is in include/linux/irqchip/orion.h and only used for .init_irq in machine descriptor later.

Quoted 7 lines
Quoted 4 lines

The irq handler is never exposed to the board file itself, but set within orion_init_irq. This approach has been taked by irqchip/irq-gic.c and irqchip/irq-vic.c rather than adding .handle_irq to the machine descriptor.

But I don't find an asm/exception.h include in drivers/irqchip/whateveryour.cfileiscalled

Well, that might be because you can magically _see_ compiler warnings while I have to actually run the compiler ;)

With the review from Jason Gunthorpe things will move anyway and I expect include/irqchip/orion.h to vanish in v2.

Sebastian

#6 lore raw
In reply to Sebastian Hesselbarth
Quoted 7 lines · Sebastian Hesselbarth
drivers/irqchip/irq-orion.c new file +6 -0
+
+static void __iomem *orion_irq_base[ORION_MAX_IRQREG];
+static unsigned int orion_irq_regs;
+static struct irq_domain *orion_irq_domain;
+
+asmlinkage void __exception_irq_entry orion_handle_irq(struct

pt_regs *regs)

This can be static?

drivers/irqchip/irq-orion.c line 97

Quoted 7 lines
drivers/irqchip/irq-orion.c new file +7 -0
+static int __init orion_of_init(struct device_node *np,
+				struct device_node *parent)
+{
+	int n;
+
+	for (n = 0; n < ORION_MAX_IRQREG; n++) {
+		orion_irq_base[n] = of_iomap(np, n);

Is it possible to also reserve the resources for these registers at this point in the boot sequence?

drivers/irqchip/irq-orion.c line 123

Quoted 3 lines
drivers/irqchip/irq-orion.c new file +3 -0
+static struct of_device_id orion_irq_dt_ids[] __initconst = {
+	{ .compatible = "marvell,orion-mpic", .data = orion_of_init },
+	{ }

Is there a strong reason to change the compatible string? Looks to me like either the new driver or the old driver will bind depending on what is in the machine description. No need for a new string?

drivers/irqchip/irq-orion.c line 129

Quoted 6 lines
drivers/irqchip/irq-orion.c new file +6 -0
+};
+
+void __init orion_init_irq(void)
+{
+	of_irq_init(orion_irq_dt_ids);
+}

Shouldn't this use the new IRQCHIP_DECLARE mechanism?

include/linux/irqchip/orion.h

Quoted 3 lines
include/linux/irqchip/orion.h +1 -0
 
+extern void orion_init_irq(void);

.. which lets this go away, use the generic irqchip_init instead of orion_init_irq.

Regards,
Jason

author maintainer
#7 lore raw
Quoted 10 lines · Jason Gunthorpe
Quoted 7 lines · Sebastian Hesselbarth
drivers/irqchip/irq-orion.c new file +6 -0
+
+static void __iomem *orion_irq_base[ORION_MAX_IRQREG];
+static unsigned int orion_irq_regs;
+static struct irq_domain *orion_irq_domain;
+
+asmlinkage void __exception_irq_entry orion_handle_irq(struct

pt_regs *regs)

This can be static?

True, corresponds with the left-over #include in linux/irqchip/orion.h.

Quoted 10 lines

drivers/irqchip/irq-orion.c line 97

Quoted 7 lines
drivers/irqchip/irq-orion.c new file +7 -0
+static int __init orion_of_init(struct device_node *np,
+				struct device_node *parent)
+{
+	int n;
+
+	for (n = 0; n < ORION_MAX_IRQREG; n++) {
+		orion_irq_base[n] = of_iomap(np, n);

Is it possible to also reserve the resources for these registers at this point in the boot sequence?

I see what I can do.

Quoted 7 lines

drivers/irqchip/irq-orion.c line 123

Quoted 3 lines
drivers/irqchip/irq-orion.c new file +3 -0
+static struct of_device_id orion_irq_dt_ids[] __initconst = {
+	{ .compatible = "marvell,orion-mpic", .data = orion_of_init },
+	{ }

Is there a strong reason to change the compatible string? Looks to me like either the new driver or the old driver will bind depending on what is in the machine description. No need for a new string?

The reason for a new compatible string is, that we will also need an secondary irq controller for bridge irqs. That could be called marvell,orion-spic. Dove is again a little bit different than the others and this will require timer and especially rtc not to share bridge irqs here. RTC irq is located in PMU regs on Dove instead of bridge regs.

But I don't have a strong opinion here and we can also reuse marvell,orion-intc for the irqchip driver.

Quoted 8 lines

drivers/irqchip/irq-orion.c line 129

Quoted 6 lines
drivers/irqchip/irq-orion.c new file +6 -0
+};
+
+void __init orion_init_irq(void)
+{
+	of_irq_init(orion_irq_dt_ids);
+}

Shouldn't this use the new IRQCHIP_DECLARE mechanism?

I didn't follow irqchip discussion lately, but will catch up.

Quoted 7 lines

include/linux/irqchip/orion.h

Quoted 4 lines
include/linux/irqchip/orion.h +1 -0
 
+extern void orion_init_irq(void);

.. which lets this go away, use the generic irqchip_init instead of orion_init_irq.

Same as above.

Thanks for the review,
Sebastian

#8 lore raw
Quoted 8 lines · Sebastian Hesselbarth
Quoted 7 lines

drivers/irqchip/irq-orion.c line 123

Quoted 3 lines
drivers/irqchip/irq-orion.c new file +3 -0
+static struct of_device_id orion_irq_dt_ids[] __initconst = {
+	{ .compatible = "marvell,orion-mpic", .data = orion_of_init },
+	{ }

Is there a strong reason to change the compatible string? Looks to me like either the new driver or the old driver will bind depending on what is in the machine description. No need for a new string?

Quoted 6 lines

The reason for a new compatible string is, that we will also need an secondary irq controller for bridge irqs. That could be called marvell,orion-spic. Dove is again a little bit different than the others and this will require timer and especially rtc not to share bridge irqs here. RTC irq is located in PMU regs on Dove instead of bridge regs.

As Arnd mentioned, I would keep the old name then..

The bridge controller can be called marvell,orion-intc-bridge, and if Dove needs a pmu controller, marvell,dove-intc-pmu ?

Quoted 4 lines
Quoted 2 lines

.. which lets this go away, use the generic irqchip_init instead of orion_init_irq.

Same as above.

I have kirkwood HW but I haven't had time to make newer kernels run on it, otherwise I'd test it too :(

Jason

author maintainer
#9 lore raw
Quoted 2 lines · Jason Gunthorpe

I have kirkwood HW but I haven't had time to make newer kernels run on it, otherwise I'd test it too :(

I also have kirkwood HW but that will cut me from email as I use it as relay server ;) Maybe I can turn it into a test bed for a while.

There is also Orion5x and Discovery Innovation (mv78xx0) to be tested.

@Jason Cooper: I will merge both irqchip and dove patches into one patch set. I wasn't earlier because I didn't want the above SoCs to slow down patch integration. And I will split dtsi changes into separate patches as requested.

Sebastian

maintainer
#10 lore raw
Quoted 8 lines · Sebastian Hesselbarth
Quoted 2 lines · Jason Gunthorpe

I have kirkwood HW but I haven't had time to make newer kernels run on it, otherwise I'd test it too :(

I also have kirkwood HW but that will cut me from email as I use it as relay server ;) Maybe I can turn it into a test bed for a while.

There is also Orion5x and Discovery Innovation (mv78xx0) to be tested.

I can test kirkwood and Orion5x.

Andrew

#11 lore raw
Quoted 11 lines · Andrew Lunn
Quoted 8 lines · Sebastian Hesselbarth
Quoted 2 lines · Jason Gunthorpe

I have kirkwood HW but I haven't had time to make newer kernels run on it, otherwise I'd test it too :(

I also have kirkwood HW but that will cut me from email as I use it as relay server ;) Maybe I can turn it into a test bed for a while.

There is also Orion5x and Discovery Innovation (mv78xx0) to be tested.

I can test kirkwood and Orion5x.

And I can test mv78xx0

Gregory

#12 lore raw
In reply to Sebastian Hesselbarth
Quoted 7 lines · Sebastian Hesselbarth
Quoted 2 lines · Jason Gunthorpe

I have kirkwood HW but I haven't had time to make newer kernels run on it, otherwise I'd test it too :(

I also have kirkwood HW but that will cut me from email as I use it as relay server ;) Maybe I can turn it into a test bed for a while.

Quoted 6 lines

There is also Orion5x and Discovery Innovation (mv78xx0) to be tested.

@Jason Cooper: I will merge both irqchip and dove patches into one patch set. I wasn't earlier because I didn't want the above SoCs to slow down patch integration. And I will split dtsi changes into separate patches as requested.

Understood. I'd prefer to keep patches developed together in the same series, that way, discussions about who is taking what and which patch depends on what are all in the same thread of the archives.

It's difficult because some series (eg pcie) touch many different areas and _need_ to be kept together because of the chain of dependencies. Most other series though, aren't in that situation.

thx,

Jason.

#13 lore raw
In reply to Jason Gunthorpe
Quoted 8 lines · Jason Gunthorpe

drivers/irqchip/irq-orion.c line 123

Quoted 3 lines
drivers/irqchip/irq-orion.c new file +3 -0
+static struct of_device_id orion_irq_dt_ids[] __initconst = {
+     { .compatible = "marvell,orion-mpic", .data = orion_of_init },
+     { }

Is there a strong reason to change the compatible string? Looks to me like either the new driver or the old driver will bind depending on what is in the machine description. No need for a new string?

The compatible string should change if the binding changes in an incomptible way, and we should try not to change it unless it's fundamentally flawed.

Arnd

author maintainer
#14 lore raw
Quoted 5 lines · Arnd Bergmann
Quoted 4 lines · Jason Gunthorpe
Quoted 3 lines
drivers/irqchip/irq-orion.c new file +3 -0
+static struct of_device_id orion_irq_dt_ids[] __initconst = {
+     { .compatible = "marvell,orion-mpic", .data = orion_of_init },
+     { }
Quoted 8 lines
Quoted 4 lines

Is there a strong reason to change the compatible string? Looks to me like either the new driver or the old driver will bind depending on what is in the machine description. No need for a new string?

The compatible string should change if the binding changes in an incomptible way, and we should try not to change it unless it's fundamentally flawed.

Well, there is no _fundamental_ change in the binding syntax as it is only reg, interrupts, and clocks. But there is a semantic change in reg properties, as current orion irq controller wants the mask registers (0x04,0x08) only while this also needs cause register (0x00).

Nothing that couldn't be handled while moving Orion arch to irqchip but if there are no further objections, I'd like to stick with the new compatible string - also having orion-spic in mind.

Sebastian

#15 lore raw
Quoted 9 lines · Sebastian Hesselbarth
Quoted 3 lines

The compatible string should change if the binding changes in an incomptible way, and we should try not to change it unless it's fundamentally flawed.

Well, there is no _fundamental_ change in the binding syntax as it is only reg, interrupts, and clocks. But there is a semantic change in reg properties, as current orion irq controller wants the mask registers (0x04,0x08) only while this also needs cause register (0x00).

Oh, I didn't notice that, good point - the original binding was flawed in that regard :|

Jason

author maintainer
#16 lore raw
In reply to Sebastian Hesselbarth
Quoted 21 lines · Sebastian Hesselbarth
Quoted 5 lines · Arnd Bergmann
Quoted 4 lines · Jason Gunthorpe
Quoted 3 lines

+static struct of_device_id orion_irq_dt_ids[] __initconst = {
+ { .compatible = "marvell,orion-mpic", .data = orion_of_init },
+ { }

Quoted 8 lines
Quoted 4 lines

Is there a strong reason to change the compatible string? Looks to me like either the new driver or the old driver will bind depending on what is in the machine description. No need for a new string?

The compatible string should change if the binding changes in an incomptible way, and we should try not to change it unless it's fundamentally flawed.

Well, there is no _fundamental_ change in the binding syntax as it is only reg, interrupts, and clocks. But there is a semantic change in reg properties, as current orion irq controller wants the mask registers (0x04,0x08) only while this also needs cause register (0x00).

Quoted 3 lines

Nothing that couldn't be handled while moving Orion arch to irqchip but if there are no further objections, I'd like to stick with the new compatible string - also having orion-spic in mind.

Taking Jason Gunthorpe's last reply, I will stick to the _old_ compatible string and name the bridge irq handler as Jason suggested.

I will send a v2 soon.

Sebastian

#17 lore raw
In reply to Sebastian Hesselbarth
Quoted 25 lines · Sebastian Hesselbarth

This patch adds an irqchip driver for the main interrupt controller found on Marvell Orion SoCs (Kirkwood, Dove, Orion5x, Discovery Innovation). Corresponding device tree documentation is also added.

---
Note: This patch triggers a checkpatch warning for
WARNING: Avoid CamelCase: <handle_IRQ>

Cc Grant Likely <[email protected]>
Cc Rob Herring <[email protected]>
Cc Rob Landley <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc Russell King <[email protected]>
Cc Arnd Bergmann <[email protected]>
Cc Jason Cooper <[email protected]>
Cc Andrew Lunn <[email protected]>
Cc Thomas Petazzoni <[email protected]>
Cc Gregory Clement <[email protected]>
Cc Ezequiel Garcia <[email protected]>
Cc Jean-Francois Moine <[email protected]>
Cc devicetree-discuss at lists.ozlabs.org
Cc linux-doc at vger.kernel.org
Cc linux-arm-kernel at lists.infradead.org
Cc linux-kernel at vger.kernel.org
Quoted 10 lines
---
 .../interrupt-controller/marvell,orion-mpic.txt    |   22 ++++
 drivers/irqchip/Kconfig                            |    5 +
 drivers/irqchip/Makefile                           |    1 +
 drivers/irqchip/irq-orion.c                        |  129 ++++++++++++++++++++
 include/linux/irqchip/orion.h                      |   18 +++
 5 files changed, 175 insertions(+), 0 deletions(-)
 create mode 100644 Documentation/devicetree/bindings/interrupt-controller/marvell,orion-mpic.txt
 create mode 100644 drivers/irqchip/irq-orion.c
 create mode 100644 include/linux/irqchip/orion.h

Sebastian,

Could you please include this patch in the next version of the series depending on it. I'd like to get an Ack from the irqchip folks for us to take this patch through mvebu/arm-soc.

This will prevent an external tree dependency, and any resulting merge conflict will be a trivial add in Kconfig or Makefile.

thx,

Jason.

Quoted 219 lines
Documentation/devicetree/bindings/interrupt-controller/marvell,orion-mpic.txt new file +22 -0
new file mode 100644
index 0000000..3b303ec
@@ -0,0 +1,22 @@
+Marvell Orion SoC main interrupt controller
+
+Required properties:
+- compatible: shall be "marvell,orion-mpic"
+- reg: base address(es) of interrupt registers starting with CAUSE register
+- interrupt-controller: identifies the node as an interrupt controller
+- #interrupt-cells: number of cells to encode an interrupt source, shall be 1.
+
+The interrupt sources map to the corresponding bits in the interrupt
+registers, i.e.
+- 0 maps to bit 0 of first base address,
+- 1 maps to bit 1 of first base address,
+- 32 maps to bit 0 of second base address, and so on.
+
+Example:
+	intc: interrupt-controller {
+		compatible = "marvell,orion-mpic";
+		interrupt-controller;
+		#interrupt-cells = <1>;
+                /* Dove has 64 primary interrupts */
+		reg = <0x20200 0x10>, <0x20210 0x10>;
+	};
drivers/irqchip/Kconfig +5 -0
index a350969..8da3559 100644
@@ -2,6 +2,11 @@ config IRQCHIP
 	def_bool y
 	depends on OF_IRQ
 
+config IRQCHIP_ORION
+	bool
+	select IRQ_DOMAIN
+	select MULTI_IRQ_HANDLER
+
 config ARM_GIC
 	bool
 	select IRQ_DOMAIN
drivers/irqchip/Makefile +1 -0
index 98e3b87..8adbd43 100644
@@ -2,6 +2,7 @@ obj-$(CONFIG_IRQCHIP)			+= irqchip.o
 
 obj-$(CONFIG_ARCH_BCM2835)		+= irq-bcm2835.o
 obj-$(CONFIG_ARCH_EXYNOS)		+= exynos-combiner.o
+obj-$(CONFIG_IRQCHIP_ORION)		+= irq-orion.o
 obj-$(CONFIG_METAG)			+= irq-metag-ext.o
 obj-$(CONFIG_METAG_PERFCOUNTER_IRQS)	+= irq-metag.o
 obj-$(CONFIG_ARCH_SUNXI)		+= irq-sunxi.o
drivers/irqchip/irq-orion.c new file +129 -0
new file mode 100644
index 0000000..ea02e11
@@ -0,0 +1,129 @@
+/*
+ * Marvell Orion SoCs IRQ chip driver.
+ *
+ * Sebastian Hesselbarth <[email protected]>
+ *
+ * This file is licensed under the terms of the GNU General Public
+ * License version 2.  This program is licensed "as is" without any
+ * warranty of any kind, whether express or implied.
+ */
+
+#include <linux/io.h>
+#include <linux/irq.h>
+#include <linux/irqchip/orion.h>
+#include <linux/of.h>
+#include <linux/of_address.h>
+#include <linux/of_irq.h>
+#include <asm/mach/irq.h>
+
+/* max number of handled irq register blocks */
+#define ORION_MAX_IRQREG		2
+
+#define ORION_IRQ_CAUSE			0x00
+#define ORION_IRQ_MASK			0x04
+#define ORION_IRQ_FIQ_MASK		0x08
+#define ORION_IRQ_ENDP_MASK		0x0c
+
+static void __iomem *orion_irq_base[ORION_MAX_IRQREG];
+static unsigned int orion_irq_regs;
+static struct irq_domain *orion_irq_domain;
+
+asmlinkage void __exception_irq_entry orion_handle_irq(struct pt_regs *regs)
+{
+	int n;
+	for (n = 0; n < orion_irq_regs; n++) {
+		u32 hwirq_base = n * 32;
+		u32 stat = readl_relaxed(orion_irq_base[n] + ORION_IRQ_CAUSE) &
+			   readl_relaxed(orion_irq_base[n] + ORION_IRQ_MASK);
+		while (stat) {
+			u32 hwirq = ffs(stat) - 1;
+			u32 irq = irq_find_mapping(orion_irq_domain,
+						hwirq_base + hwirq);
+			handle_IRQ(irq, regs);
+			stat &= ~(1 << hwirq);
+		}
+	}
+}
+
+static void orion_irq_mask(struct irq_data *irqd)
+{
+	unsigned int irq = irqd_to_hwirq(irqd);
+	unsigned int irq_off = irq % 32;
+	int reg = irq / 32;
+	u32 val;
+
+	val = readl(orion_irq_base[reg] + ORION_IRQ_MASK);
+	writel(val & ~(1 << irq_off), orion_irq_base[reg] + ORION_IRQ_MASK);
+}
+
+static void orion_irq_unmask(struct irq_data *irqd)
+{
+	unsigned int irq = irqd_to_hwirq(irqd);
+	unsigned int irq_off = irq % 32;
+	int reg = irq / 32;
+	u32 val;
+
+	val = readl(orion_irq_base[reg] + ORION_IRQ_MASK);
+	writel(val | (1 << irq_off), orion_irq_base[reg] + ORION_IRQ_MASK);
+}
+
+static struct irq_chip orion_irq_chip = {
+	.name		= "orion_irq",
+	.irq_mask	= orion_irq_mask,
+	.irq_unmask	= orion_irq_unmask,
+};
+
+static int orion_irq_map(struct irq_domain *d, unsigned int virq,
+			 irq_hw_number_t hw)
+{
+	irq_set_chip_and_handler(virq, &orion_irq_chip,
+				 handle_level_irq);
+	set_irq_flags(virq, IRQF_VALID | IRQF_PROBE);
+
+	return 0;
+}
+
+static struct irq_domain_ops orion_irq_ops = {
+	.map = orion_irq_map,
+	.xlate = irq_domain_xlate_onecell,
+};
+
+static int __init orion_of_init(struct device_node *np,
+				struct device_node *parent)
+{
+	int n;
+
+	for (n = 0; n < ORION_MAX_IRQREG; n++) {
+		orion_irq_base[n] = of_iomap(np, n);
+
+		if (!orion_irq_base[n])
+			continue;
+
+		/* mask all interrupts */
+		writel(0, orion_irq_base[n] + ORION_IRQ_MASK);
+		orion_irq_regs++;
+	}
+
+	/* at least one irq reg must be set */
+	if (!orion_irq_regs)
+		panic("%s: unable to map IRQC registers\n", np->full_name);
+
+	orion_irq_domain = irq_domain_add_linear(np, orion_irq_regs * 32,
+						 &orion_irq_ops, NULL);
+	if (!orion_irq_domain)
+		panic("%s: unable to create IRQ domain\n", np->full_name);
+
+	set_handle_irq(orion_handle_irq);
+
+	return 0;
+}
+
+static struct of_device_id orion_irq_dt_ids[] __initconst = {
+	{ .compatible = "marvell,orion-mpic", .data = orion_of_init },
+	{ }
+};
+
+void __init orion_init_irq(void)
+{
+	of_irq_init(orion_irq_dt_ids);
+}
include/linux/irqchip/orion.h new file +18 -0
new file mode 100644
index 0000000..04f7bab
@@ -0,0 +1,18 @@
+/*
+ * Marvell Orion SoCs IRQ chip driver header.
+ *
+ * Sebastian Hesselbarth <[email protected]>
+ *
+ * This file is licensed under the terms of the GNU General Public
+ * License version 2.  This program is licensed "as is" without any
+ * warranty of any kind, whether express or implied.
+ */
+
+#ifndef __LINUX_IRQCHIP_ORION_H
+#define __LINUX_IRQCHIP_ORION_H
+
+#include <asm/exception.h>
+
+extern void orion_init_irq(void);
+
+#endif
Signature · 1 line
1.7.2.5
#18 lore raw
In reply to Sebastian Hesselbarth

Sebastian,

please do not take the rant below personally. You just happen to trigger it.

Quoted 11 lines · Sebastian Hesselbarth
drivers/irqchip/irq-orion.c new file +11 -0
+static void orion_irq_mask(struct irq_data *irqd)
+{
+	unsigned int irq = irqd_to_hwirq(irqd);
+	unsigned int irq_off = irq % 32;
+	int reg = irq / 32;
+	u32 val;
+
+	val = readl(orion_irq_base[reg] + ORION_IRQ_MASK);
+	writel(val & ~(1 << irq_off), orion_irq_base[reg] + ORION_IRQ_MASK);
+}
+

drivers/irqchip/irq-orion.c line 68

Quoted 10 lines
drivers/irqchip/irq-orion.c new file +10 -0
+static void orion_irq_unmask(struct irq_data *irqd)
+{
+	unsigned int irq = irqd_to_hwirq(irqd);
+	unsigned int irq_off = irq % 32;
+	int reg = irq / 32;
+	u32 val;
+
+	val = readl(orion_irq_base[reg] + ORION_IRQ_MASK);
+	writel(val | (1 << irq_off), orion_irq_base[reg] + ORION_IRQ_MASK);
+}

I'm really tired of looking at the next incarnation of an OF/DT irq chip driver, which reimplements stuff which I have consolidated in the generic irq chip implementation with a lot of effort.

Just look at the various implementations in drivers/irqchip/ and find out how similar they are. Moving code to drivers/irqchip/ does not make an excuse for reestablishing the mess which was addressed by the generic irq chip implementation.

Can you - and that means all of you ARM folks - please get your gear together and add the missing features to the generic irq chip implementation? I'm not going to accept more of that OF/DT frenzy.

Thanks,

tglx

author maintainer
#19 lore raw
Quoted 2 lines · Thomas Gleixner

please do not take the rant below personally. You just happen to trigger it.

Thomas,

it is okay for me - but thanks for the notice! I will comment below.

Quoted 18 lines · Sebastian Hesselbarth
Quoted 17 lines
drivers/irqchip/irq-orion.c new file +17 -0
+static void orion_irq_mask(struct irq_data *irqd)
+{
+	unsigned int irq = irqd_to_hwirq(irqd);
+	unsigned int irq_off = irq % 32;
+	int reg = irq / 32;
+	u32 val;
+
+	val = readl(orion_irq_base[reg] + ORION_IRQ_MASK);
+	writel(val&  ~(1<<  irq_off), orion_irq_base[reg] + ORION_IRQ_MASK);
+}
+
+static void orion_irq_unmask(struct irq_data *irqd)
+{
+	unsigned int irq = irqd_to_hwirq(irqd);
+	unsigned int irq_off = irq % 32;
+	int reg = irq / 32;
+	u32 val;
Quoted 8 lines

drivers/irqchip/irq-orion.c line 57

Quoted 4 lines
drivers/irqchip/irq-orion.c new file +4 -0
+
+	val = readl(orion_irq_base[reg] + ORION_IRQ_MASK);
+	writel(val | (1<<  irq_off), orion_irq_base[reg] + ORION_IRQ_MASK);
+}

I'm really tired of looking at the next incarnation of an OF/DT irq chip driver, which reimplements stuff which I have consolidated in the generic irq chip implementation with a lot of effort.

Actually, non-irqchip implementation of orion intc was based on generic irq chip already. I took a look at drivers/irqchip and realized that at least sunxi (ARM again) was reimplementing mask/unmask the way above. So I took the short path and copied that.

Quoted 8 lines

Just look at the various implementations in drivers/irqchip/ and find out how similar they are. Moving code to drivers/irqchip/ does not make an excuse for reestablishing the mess which was addressed by the generic irq chip implementation.

Can you - and that means all of you ARM folks - please get your gear together and add the missing features to the generic irq chip implementation? I'm not going to accept more of that OF/DT frenzy.

So you are suggesting to have a "linux,generic-intc" or you want me to have "marvell,orion-intc" make use of generic irq chip again?

The second is easy, the first will take me a while to think about proper DT properties how to encode mask/unmask/ack/.. availability and offsets.

Regards,
Sebastian

#20 lore raw
Quoted 9 lines · Sebastian Hesselbarth
Quoted 8 lines

Just look at the various implementations in drivers/irqchip/ and find out how similar they are. Moving code to drivers/irqchip/ does not make an excuse for reestablishing the mess which was addressed by the generic irq chip implementation.

Can you - and that means all of you ARM folks - please get your gear together and add the missing features to the generic irq chip implementation? I'm not going to accept more of that OF/DT frenzy.

Quoted 6 lines

So you are suggesting to have a "linux,generic-intc" or you want me to have "marvell,orion-intc" make use of generic irq chip again?

The second is easy, the first will take me a while to think about proper DT properties how to encode mask/unmask/ack/.. availability and offsets.

I think it should not be "linux,..." since the description can well be OS independent. I don't have a good idea for a generic name, but it's not very important.

The main missing piece in the generic irqchip code at the moment is support for the linear irqdomain. Once we have that, a lot of code can be simplified significantly.

Arnd

author maintainer
#21 lore raw
Quoted 17 lines · Arnd Bergmann
Quoted 9 lines · Sebastian Hesselbarth
Quoted 8 lines

Just look at the various implementations in drivers/irqchip/ and find out how similar they are. Moving code to drivers/irqchip/ does not make an excuse for reestablishing the mess which was addressed by the generic irq chip implementation.

Can you - and that means all of you ARM folks - please get your gear together and add the missing features to the generic irq chip implementation? I'm not going to accept more of that OF/DT frenzy.

Quoted 6 lines

So you are suggesting to have a "linux,generic-intc" or you want me to have "marvell,orion-intc" make use of generic irq chip again?

The second is easy, the first will take me a while to think about proper DT properties how to encode mask/unmask/ack/.. availability and offsets.

Quoted 7 lines

I think it should not be "linux,..." since the description can well be OS independent. I don't have a good idea for a generic name, but it's not very important.

The main missing piece in the generic irqchip code at the moment is support for the linear irqdomain. Once we have that, a lot of code can be simplified significantly.

Arnd, Thomas,

I still don't get it completely. Are you requesting a full blown DT-enabled generic irq chip driver that can be setup by DT properties?

Or is it just to have current generic irq chip play with linear irqdomain?

If it is just linear irqdomain, then I can have a look at it now. I would first send v2 to allow the others to integrate Kirkwood, Orion5x, and Discovery Innovation. Then v3 will touch generic irq and merge that with orion irqchip.

(@Jason C: Are you sure that I should merge dove and orion irqchip patches? I doubt that anything touching generic irq will not go through irq tree.)

Sebastian

#22 lore raw
Quoted 3 lines · Sebastian Hesselbarth

(@Jason C: Are you sure that I should merge dove and orion irqchip patches? I doubt that anything touching generic irq will not go through irq tree.)

Putting them in the same patch series does not imply they have to go through the same tree. But it *does* allow us to see relationships, conflicts, etc.

Based on how the finally dependencies work out, we may ask the irq maintainers for an Ack to take it through arm-soc. This would happen if we can't remove the dependency between the trees. The resulting potential merge conflicts weigh into it, and that's where the irqchip maintainer's Ack get decided.

If the changes to irqchip are just Makefile/Kconfig, then it's easy. However, if several other files are changed and conflicting, then we let it go through irqchip and wait one merge window for the board changes depending on it.

The goal here is to identify and remove branch dependencies within arm-soc and between arm-soc and other trees. A secondary goal is to identify high-risk series (risk of being dropped), and keep them in separate branches from other changes.

thx,

Jason.

author maintainer
#23 lore raw

v2 cover ARM: orion: add orion irqchip driver

In reply to Sebastian Hesselbarth

This patch set adds an irqchip driver for the main interrupt controller found on Marvell Orion SoCs (Kirkwood, Dove, Orion5x, Discovery Innovation) and the corresponding device tree documentation.

It also moves Dove as the first Orion SoC to this irqchip driver. As legacy non-DT drivers don't know about virtual irqs provided by irqchip drivers, init code for timer and mv643xx_eth is added until true DT enabled drivers are available. Finally, DT enabled boards for Dove are moved to the orion irqchip driver.

*Note:*
This patch set is _not_ feature complete. There has been a request for irqdomain support for generic irq and perhaps even a DT enabled generic irq chip driver.

The patch set still can be used to port other Orion SoCs as I did with Dove. It also merges irqchip driver patches with Dove patches to use this driver.

Sebastian Hesselbarth (5):
irqchip: add support for Marvell Orion SoCs
ARM: dove: add DT parsing for legacy mv643xx_eth
ARM: dove: add DT parsing for legacy timer
ARM: dove: move DT boards to orion irqchip driver
ARM: dove: add DT nodes for irqchip conversion

 .../interrupt-controller/marvell,orion-intc.txt    |   22 ++++
 arch/arm/boot/dts/dove.dtsi                        |   16 ++-
 arch/arm/mach-dove/Kconfig                         |    1 +
 arch/arm/mach-dove/Makefile                        |    4 +-
 arch/arm/mach-dove/board-dt.c                      |   77 ++++++++++--
 drivers/irqchip/Kconfig                            |    5 +
 drivers/irqchip/Makefile                           |    1 +
 drivers/irqchip/irq-orion.c                        |  133 ++++++++++++++++++++
 8 files changed, 247 insertions(+), 12 deletions(-)
 create mode 100644 Documentation/devicetree/bindings/interrupt-controller/marvell,orion-intc.txt
 create mode 100644 drivers/irqchip/irq-orion.c

---
Cc: Grant Likely <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: Rob Landley <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Russell King <[email protected]>
Cc: Arnd Bergmann <[email protected]>
Cc: Jason Cooper <[email protected]>
Cc: Andrew Lunn <[email protected]>
Cc: Jason Gunthorpe <[email protected]>
Cc: Thomas Petazzoni <[email protected]>
Cc: Gregory Clement <[email protected]>
Cc: Ezequiel Garcia <[email protected]>
Cc: Jean-Francois Moine <[email protected]>
Cc: devicetree-discuss at lists.ozlabs.org
Cc: linux-doc at vger.kernel.org
Cc: linux-arm-kernel at lists.infradead.org
Cc: linux-kernel at vger.kernel.org

Signature · 1 line
1.7.10.4
author maintainer
#24 lore raw

PATCH v2 1/5 irqchip: add support for Marvell Orion SoCs

This patch adds an irqchip driver for the main interrupt controller found on Marvell Orion SoCs (Kirkwood, Dove, Orion5x, Discovery Innovation). Corresponding device tree documentation is also added.

---
Note: This patch triggers a checkpatch warning for
WARNING: Avoid CamelCase: <handle_IRQ>

Changelog:
v1->v2:
- rename compatible string to "marvell,orion-intc" (Suggested by Jason Gunthorpe)
- request mem regions for irq base registers (Suggested by Jason Gunthorpe)
- make orion_handle_irq static (Suggested by Jason Gunthorpe)
- make use of IRQCHIP_DECLARE (Suggested by Jason Gunthorpe)

Cc Grant Likely <[email protected]>
Cc Rob Herring <[email protected]>
Cc Rob Landley <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc Russell King <[email protected]>
Cc Arnd Bergmann <[email protected]>
Cc Jason Cooper <[email protected]>
Cc Andrew Lunn <[email protected]>
Cc Jason Gunthorpe <[email protected]>
Cc Thomas Petazzoni <[email protected]>
Cc Gregory Clement <[email protected]>
Cc Ezequiel Garcia <[email protected]>
Cc Jean-Francois Moine <[email protected]>
Cc devicetree-discuss at lists.ozlabs.org
Cc linux-doc at vger.kernel.org
Cc linux-arm-kernel at lists.infradead.org
Cc linux-kernel at vger.kernel.org
---
 .../interrupt-controller/marvell,orion-intc.txt    |   22 ++++
 drivers/irqchip/Kconfig                            |    5 +
 drivers/irqchip/Makefile                           |    1 +
 drivers/irqchip/irq-orion.c                        |  133 ++++++++++++++++++++
 4 files changed, 161 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/interrupt-controller/marvell,orion-intc.txt
 create mode 100644 drivers/irqchip/irq-orion.c
Documentation/devicetree/bindings/interrupt-controller/marvell,orion-intc.txt new file +22 -0
new file mode 100644
index 0000000..9b7aee9
@@ -0,0 +1,22 @@
+Marvell Orion SoC main interrupt controller
+
+Required properties:
+- compatible: shall be "marvell,orion-intc"
+- reg: base address(es) of interrupt registers starting with CAUSE register
+- interrupt-controller: identifies the node as an interrupt controller
+- #interrupt-cells: number of cells to encode an interrupt source, shall be 1.
+
+The interrupt sources map to the corresponding bits in the interrupt
+registers, i.e.
+- 0 maps to bit 0 of first base address,
+- 1 maps to bit 1 of first base address,
+- 32 maps to bit 0 of second base address, and so on.
+
+Example:
+	intc: interrupt-controller {
+		compatible = "marvell,orion-intc";
+		interrupt-controller;
+		#interrupt-cells = <1>;
+                /* Dove has 64 first level interrupts */
+		reg = <0x20200 0x10>, <0x20210 0x10>;
+	};
drivers/irqchip/Kconfig +5 -0
index a350969..8da3559 100644
@@ -2,6 +2,11 @@ config IRQCHIP
 	def_bool y
 	depends on OF_IRQ
 
+config IRQCHIP_ORION
+	bool
+	select IRQ_DOMAIN
+	select MULTI_IRQ_HANDLER
+
 config ARM_GIC
 	bool
 	select IRQ_DOMAIN
drivers/irqchip/Makefile +1 -0
index 10ef57f..2cad23d 100644
@@ -5,6 +5,7 @@ obj-$(CONFIG_ARCH_EXYNOS)		+= exynos-combiner.o
 obj-$(CONFIG_ARCH_MXS)			+= irq-mxs.o
 obj-$(CONFIG_METAG)			+= irq-metag-ext.o
 obj-$(CONFIG_METAG_PERFCOUNTER_IRQS)	+= irq-metag.o
+obj-$(CONFIG_IRQCHIP_ORION)		+= irq-orion.o
 obj-$(CONFIG_ARCH_SUNXI)		+= irq-sun4i.o
 obj-$(CONFIG_ARCH_SPEAR3XX)		+= spear-shirq.o
 obj-$(CONFIG_ARM_GIC)			+= irq-gic.o
drivers/irqchip/irq-orion.c new file +133 -0

133 lines not on this page · raw message

Signature · 1 line
1.7.10.4
#25 lore raw
Quoted 9 lines · Sebastian Hesselbarth

This patch adds an irqchip driver for the main interrupt controller found on Marvell Orion SoCs (Kirkwood, Dove, Orion5x, Discovery Innovation). Corresponding device tree documentation is also added.

---
Note: This patch triggers a checkpatch warning for
WARNING: Avoid CamelCase: <handle_IRQ>

Quoted 6 lines

Changelog:
v1->v2:
- rename compatible string to "marvell,orion-intc" (Suggested by Jason Gunthorpe)
- request mem regions for irq base registers (Suggested by Jason Gunthorpe)
- make orion_handle_irq static (Suggested by Jason Gunthorpe)
- make use of IRQCHIP_DECLARE (Suggested by Jason Gunthorpe)

It would still be a good idea to convert this to use the generic chip stuff which Thomas created, though exactly how is hard to see at the moment.

drivers/irqchip/irq-orion.c line 60

Quoted 10 lines
drivers/irqchip/irq-orion.c new file +10 -0
+static void orion_irq_mask(struct irq_data *irqd)
+{
+	unsigned int irq = irqd_to_hwirq(irqd);
+	unsigned int irq_off = irq % 32;
+	int reg = irq / 32;
+	u32 val;
+
+	val = readl(orion_irq_base[reg] + ORION_IRQ_MASK);
+	writel(val & ~(1 << irq_off), orion_irq_base[reg] + ORION_IRQ_MASK);
+}

This could be replaced with irq_gc_mask_clr_bit().

drivers/irqchip/irq-orion.c line 71

Quoted 11 lines
drivers/irqchip/irq-orion.c new file +11 -0
+
+static void orion_irq_unmask(struct irq_data *irqd)
+{
+	unsigned int irq = irqd_to_hwirq(irqd);
+	unsigned int irq_off = irq % 32;
+	int reg = irq / 32;
+	u32 val;
+
+	val = readl(orion_irq_base[reg] + ORION_IRQ_MASK);
+	writel(val | (1 << irq_off), orion_irq_base[reg] + ORION_IRQ_MASK);
+}

This with irq_gc_mask_set_bit().

drivers/irqchip/irq-orion.c line 84

Quoted 13 lines
drivers/irqchip/irq-orion.c new file +13 -0
+
+static struct irq_chip orion_irq_chip = {
+	.name		= "orion_irq",
+	.irq_mask	= orion_irq_mask,
+	.irq_unmask	= orion_irq_unmask,
+};
+
+static int orion_irq_map(struct irq_domain *d, unsigned int virq,
+			 irq_hw_number_t hw)
+{
+	irq_set_chip_and_handler(virq, &orion_irq_chip,
+				 handle_level_irq);
+	set_irq_flags(virq, IRQF_VALID | IRQF_PROBE);

This is where it starts to get tricky, because I can't see how you'd merge the irq_alloc_generic_chip() and irq_setup_generic_chip() with this. Maybe you don't need to do anything here and just do all that in orion_of_init() instead? But then you seem to need to know the virq range before hand, and I can't see how that's known. Maybe Thomas can provide some enlightenment about how the gc irqchip stuff works with the irq domain stuff...

However, you wouldn't need the statically defined orion_irq_chip nor orion_irq_base[] either, because those would be held within the gc irqchip stuff and stored in the upper layer.

author maintainer
#26 lore raw
Quoted 17 lines · Russell King - ARM Linux
Quoted 9 lines · Sebastian Hesselbarth

This patch adds an irqchip driver for the main interrupt controller found on Marvell Orion SoCs (Kirkwood, Dove, Orion5x, Discovery Innovation). Corresponding device tree documentation is also added.

---
Note: This patch triggers a checkpatch warning for
WARNING: Avoid CamelCase: <handle_IRQ>

Quoted 6 lines

Changelog:
v1->v2:
- rename compatible string to "marvell,orion-intc" (Suggested by Jason Gunthorpe)
- request mem regions for irq base registers (Suggested by Jason Gunthorpe)
- make orion_handle_irq static (Suggested by Jason Gunthorpe)
- make use of IRQCHIP_DECLARE (Suggested by Jason Gunthorpe)

Quoted 3 lines

It would still be a good idea to convert this to use the generic chip stuff which Thomas created, though exactly how is hard to see at the moment.

Russel,

that is the plan and that's why the whole patch set is preliminary. Maybe it would have been more precise to call it RFC instead.

Quoted 41 lines

drivers/irqchip/irq-orion.c line 60

Quoted 10 lines
drivers/irqchip/irq-orion.c new file +10 -0
+static void orion_irq_mask(struct irq_data *irqd)
+{
+	unsigned int irq = irqd_to_hwirq(irqd);
+	unsigned int irq_off = irq % 32;
+	int reg = irq / 32;
+	u32 val;
+
+	val = readl(orion_irq_base[reg] + ORION_IRQ_MASK);
+	writel(val & ~(1 << irq_off), orion_irq_base[reg] + ORION_IRQ_MASK);
+}

This could be replaced with irq_gc_mask_clr_bit().

drivers/irqchip/irq-orion.c line 71

Quoted 11 lines
drivers/irqchip/irq-orion.c new file +11 -0
+
+static void orion_irq_unmask(struct irq_data *irqd)
+{
+	unsigned int irq = irqd_to_hwirq(irqd);
+	unsigned int irq_off = irq % 32;
+	int reg = irq / 32;
+	u32 val;
+
+	val = readl(orion_irq_base[reg] + ORION_IRQ_MASK);
+	writel(val | (1 << irq_off), orion_irq_base[reg] + ORION_IRQ_MASK);
+}

This with irq_gc_mask_set_bit().

drivers/irqchip/irq-orion.c line 84

Quoted 13 lines
drivers/irqchip/irq-orion.c new file +13 -0
+
+static struct irq_chip orion_irq_chip = {
+	.name		= "orion_irq",
+	.irq_mask	= orion_irq_mask,
+	.irq_unmask	= orion_irq_unmask,
+};
+
+static int orion_irq_map(struct irq_domain *d, unsigned int virq,
+			 irq_hw_number_t hw)
+{
+	irq_set_chip_and_handler(virq, &orion_irq_chip,
+				 handle_level_irq);
+	set_irq_flags(virq, IRQF_VALID | IRQF_PROBE);
Quoted 7 lines

This is where it starts to get tricky, because I can't see how you'd merge the irq_alloc_generic_chip() and irq_setup_generic_chip() with this. Maybe you don't need to do anything here and just do all that in orion_of_init() instead? But then you seem to need to know the virq range before hand, and I can't see how that's known. Maybe Thomas can provide some enlightenment about how the gc irqchip stuff works with the irq domain stuff...

Exactly, and that is what I am looking into right now. But hell, I am not an expert in linux irq yet. Moreover, I am not even sure if it is okay to rely on irqdomain or at least irq_data->hw_irq at all.

My current impression is, that generic chip knowns nothing about irq domains. But my first modification of it was to use irqd_to_hwirq(d) where ever it uses d->irq instead. This should allow to abstract from virtual irqs and retain compatibility (_if_ hw_irq is also set on !CONFIG_IRQ_DOMAIN).

To add more juice: IRQF_VALID and IRQF_PROBE are ARM only flags. I tried to find out what they are good for, but stopped googl'ing after a while. (I know you explained that before somewhere)

Quoted 3 lines

However, you wouldn't need the statically defined orion_irq_chip nor orion_irq_base[] either, because those would be held within the gc irqchip stuff and stored in the upper layer.

Yeah, that would be very nice. But the current limitation to one register set with max 32 irqs of generic chip would still require to keep a list of primary generic irq chips to flip through in the irq_handler.

This also raises the question, how to check if an generic irq chip flow handler has to be called. Current irq_chip_regs don't know nothing about a cause/status register. And actually you don't even know if it is high/low active and how to mask it with the high/low active mask register or mask_cache.

Sebastian

#27 lore raw
Quoted 9 lines · Sebastian Hesselbarth
Quoted 7 lines · Russell King - ARM Linux

This is where it starts to get tricky, because I can't see how you'd merge the irq_alloc_generic_chip() and irq_setup_generic_chip() with this. Maybe you don't need to do anything here and just do all that in orion_of_init() instead? But then you seem to need to know the virq range before hand, and I can't see how that's known. Maybe Thomas can provide some enlightenment about how the gc irqchip stuff works with the irq domain stuff...

Quoted 3 lines

Exactly, and that is what I am looking into right now. But hell, I am not an expert in linux irq yet. Moreover, I am not even sure if it is okay to rely on irqdomain or at least irq_data->hw_irq at all.

Here is a solution to that problem.

1) It adds a mask field to irq_data so we dont have to compute the mask over and over

2) For compability with existing users it populates the mask with 1 << (d->irq - gc->irq_base)

3) It gives you the option to disable that mask setup or let it generate from d->hwirq

I'm still looking into a way how to proper support the generic chip / linear domain mapping in the setup path. Will send you a draft patch to play with later.

Thanks,

tglx

include/linux/irq.h +6 -0
Index: linux-2.6/include/linux/irq.h
===================================================================
@@ -119,6 +119,7 @@ struct irq_domain;
 
 /**
  * struct irq_data - per irq and irq chip data passed down to chip functions
+ * @mask:		precomputed bitmask for accessing the chip registers
  * @irq:		interrupt number
  * @hwirq:		hardware interrupt number, local to the interrupt domain
  * @node:		node index useful for balancing
@@ -138,6 +139,7 @@ struct irq_domain;
  * irq_data.
  */
 struct irq_data {
+	u32			mask;
 	unsigned int		irq;
 	unsigned long		hwirq;
 	unsigned int		node;
@@ -700,10 +702,14 @@ struct irq_chip_generic {
  * @IRQ_GC_INIT_NESTED_LOCK:	Set the lock class of the irqs to nested for
  *				irq chips which need to call irq_set_wake() on
  *				the parent irq. Usually GPIO implementations
+ * @IRQ_GC_NO_MASK:		Do not calculate irq_data->mask
+ * @IRQ_GC_MASK_FROM_HWIRQ:	Calculate irq_data->mask from the hwirq number
  */
 enum irq_gc_flags {
 	IRQ_GC_INIT_MASK_CACHE		= 1 << 0,
 	IRQ_GC_INIT_NESTED_LOCK		= 1 << 1,
+	IRQ_GC_NO_MASK			= 1 << 2,
+	IRQ_GC_MASK_FROM_HWIRQ		= 1 << 4,
 };
 
 /* Generic chip callback functions */
kernel/irq/generic-chip.c +18 -9
Index: linux-2.6/kernel/irq/generic-chip.c
===================================================================
@@ -39,7 +39,7 @@ void irq_gc_noop(struct irq_data *d)
 void irq_gc_mask_disable_reg(struct irq_data *d)
 {
 	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
-	u32 mask = 1 << (d->irq - gc->irq_base);
+	u32 mask = d->mask;
 
 	irq_gc_lock(gc);
 	irq_reg_writel(mask, gc->reg_base + cur_regs(d)->disable);
@@ -57,7 +57,7 @@ void irq_gc_mask_disable_reg(struct irq_
 void irq_gc_mask_set_bit(struct irq_data *d)
 {
 	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
-	u32 mask = 1 << (d->irq - gc->irq_base);
+	u32 mask = d->mask;
 
 	irq_gc_lock(gc);
 	gc->mask_cache |= mask;
@@ -75,7 +75,7 @@ void irq_gc_mask_set_bit(struct irq_data
 void irq_gc_mask_clr_bit(struct irq_data *d)
 {
 	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
-	u32 mask = 1 << (d->irq - gc->irq_base);
+	u32 mask = d->mask;
 
 	irq_gc_lock(gc);
 	gc->mask_cache &= ~mask;
@@ -93,7 +93,7 @@ void irq_gc_mask_clr_bit(struct irq_data
 void irq_gc_unmask_enable_reg(struct irq_data *d)
 {
 	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
-	u32 mask = 1 << (d->irq - gc->irq_base);
+	u32 mask = d->mask;
 
 	irq_gc_lock(gc);
 	irq_reg_writel(mask, gc->reg_base + cur_regs(d)->enable);
@@ -108,7 +108,7 @@ void irq_gc_unmask_enable_reg(struct irq
 void irq_gc_ack_set_bit(struct irq_data *d)
 {
 	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
-	u32 mask = 1 << (d->irq - gc->irq_base);
+	u32 mask = d->mask;
 
 	irq_gc_lock(gc);
 	irq_reg_writel(mask, gc->reg_base + cur_regs(d)->ack);
@@ -122,7 +122,7 @@ void irq_gc_ack_set_bit(struct irq_data 
 void irq_gc_ack_clr_bit(struct irq_data *d)
 {
 	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
-	u32 mask = ~(1 << (d->irq - gc->irq_base));
+	u32 mask = ~d->mask;
 
 	irq_gc_lock(gc);
 	irq_reg_writel(mask, gc->reg_base + cur_regs(d)->ack);
@@ -136,7 +136,7 @@ void irq_gc_ack_clr_bit(struct irq_data 
 void irq_gc_mask_disable_reg_and_ack(struct irq_data *d)
 {
 	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
-	u32 mask = 1 << (d->irq - gc->irq_base);
+	u32 mask = d->mask;
 
 	irq_gc_lock(gc);
 	irq_reg_writel(mask, gc->reg_base + cur_regs(d)->mask);
@@ -151,7 +151,7 @@ void irq_gc_mask_disable_reg_and_ack(str
 void irq_gc_eoi(struct irq_data *d)
 {
 	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
-	u32 mask = 1 << (d->irq - gc->irq_base);
+	u32 mask = d->mask;
 
 	irq_gc_lock(gc);
 	irq_reg_writel(mask, gc->reg_base + cur_regs(d)->eoi);
@@ -169,7 +169,7 @@ void irq_gc_eoi(struct irq_data *d)
 int irq_gc_set_wake(struct irq_data *d, unsigned int on)
 {
 	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
-	u32 mask = 1 << (d->irq - gc->irq_base);
+	u32 mask = d->mask;
 
 	if (!(mask & gc->wake_enabled))
 		return -EINVAL;
@@ -254,6 +254,15 @@ void irq_setup_generic_chip(struct irq_c
 		if (flags & IRQ_GC_INIT_NESTED_LOCK)
 			irq_set_lockdep_class(i, &irq_nested_lock_class);
 
+		if (!(flags & IRQ_GC_NO_MASK)) {
+			struct irq_data *d = irq_get_irq_data(i);
+			u32 mask;
+
+			if (flags & IRQ_GC_MASK_FROM_HWIRQ)
+				d->mask = 1 << (d->hwirq % 32);
+			else
+				d->mask = 1 << (i - gc->irq_base);
+		}
 		irq_set_chip_and_handler(i, &ct->chip, ct->handler);
 		irq_set_chip_data(i, gc);
 		irq_modify_status(i, clr, set);
#28 lore raw

cover genirq: Support for irq domains in generic irq chip

The ongoing device tree support for ARM is creating new irq chip drivers in drivers/irqchip/ in a frenzy. Quite some of them are ripping out the generic irq chip implementation from arch/arm/* and just creating the same mess of duplicated code again, which was cleaned up with the generic irq chip implementation with a lot of effort. Sigh!

I already prodded a few people in reviews to tackle that issue with no outcome. Even more sigh!

Poor Sebastian triggered me into rant mode, but he ad hoc volunteered to give it a try. YAY!

Though he asked for a bit of kickstart help. So I squeezed out a few spare cycles and implemented the basics as far as I think that they should work.

The following series contains the missing bits and pieces including a somehow forgotten and now slightly modified series from Gerlando adding support for irq chips which need separate mask caches for different chip (control flow) types.

At the moment this supports only linear irq domains, but it could be extended to other types as well if the need arises. Though the ARM chips are pretty much all about linear domains AFAICT.

It also lacks support for removing an irq domain at the moment, but that should be rather trivial to fix.

The last patch in the series is a blind conversion of the irq-sun4i irq chip driver, completely untested and not even compiled. I just added it for demonstration purposes. As Russell expected, there is a lot of consolidation potential. The changelog of that patch is:

 1 file changed, 29 insertions(+), 71 deletions(-)

The preparing series has

 4 files changed, 294 insertions(+), 50 deletions(-)

So for removing 42 lines in a single driver the core grows 244 lines including header changes and comments. Convert 6 drivers and we are more than even because we get the benefit of sharing and therefor exposing the same code to broader testing and utilization.

We have already 11 of those candidates in drivers/irqchips and new ones are knocking on the door.

There might be even more consolidation potential, but I leave that to the DT/irq domain experts.

WARNING: It's compile tested only. So if you find bugs you can keep them and fix them yourself :)

Thanks,

tglx

---
 drivers/irqchip/irq-sun4i.c |  100 ++++-----------
 include/linux/irq.h         |   45 ++++++-
 include/linux/irqdomain.h   |   12 +
 kernel/irq/generic-chip.c   |  281 +++++++++++++++++++++++++++++++++++++-------
 kernel/irq/irqdomain.c      |    6 
 5 files changed, 323 insertions(+), 121 deletions(-)
#30 lore raw

RFC PATCH 1/8 genirq: generic chip: Remove the local cur_regs() function

Quoted 5 lines · Thomas Gleixner

From: Gerlando Falauto <[email protected]>

Since we already have an irq_data_get_chip_type() function which returns a pointer to irq_chip_type, use that instead of cur_regs().

Quoted 11 lines
Cc Lennert Buytenhek <[email protected]>
Cc Simon Guinot <[email protected]>
Cc Joey Oravec <[email protected]>
Cc Ben Dooks <[email protected]>
Cc Nicolas Pitre <[email protected]>
Cc Jason Cooper <[email protected]>
Cc Andrew Lunn <[email protected]>
Cc Holger Brunck <[email protected]>
Cc linux-arm-kernel at lists.infradead.org

g.

#34 lore raw

Re: [RFC patch 4/8] genirq: generic chip: Cache per irq bit mask

Review

Quoted 2 lines · Thomas Gleixner
(quoted diff) +1 -1
-	u32 mask = ~(1 << (d->irq - gc->irq_base));
+	u32 mask = ~(d->mask);

I suspect the above changes are all a result of a search-and-replace which results in the cosmetic parens remaining. Would be nice to get rid of them too as they're completely unnecessary.

#35 lore raw

Re: [RFC patch 4/8] genirq: generic chip: Cache per irq bit mask

Quoted 7 lines · Russell King - ARM Linux

Review

Quoted 2 lines · Thomas Gleixner
(quoted diff) +1 -1
-	u32 mask = ~(1 << (d->irq - gc->irq_base));
+	u32 mask = ~(d->mask);

I suspect the above changes are all a result of a search-and-replace which results in the cosmetic parens remaining. Would be nice to get rid of them too as they're completely unnecessary.

Fair enough.

Thanks,

tglx

#40 lore raw

Re: [RFC patch 7/8] genirq: generic chip: Add linear irq domain support

Review

Quoted 6 lines · Thomas Gleixner
(quoted diff) +6 -0
+	/* Init mask cache ? */
+	if (dgc->gc_flags & IRQ_GC_INIT_MASK_CACHE) {
+		raw_spin_lock_irqsave(&gc->lock, flags);
+		gc->mask_cache = irq_reg_readl(gc->reg_base + ct->regs.mask);
+		raw_spin_unlock_irqrestore(&gc->lock, flags);
+	}

This looks a little weird to me - it seems that it'll re-read this each time any irq is mapped in the domain, which is probably not wanted.

#41 lore raw

Re: [RFC patch 7/8] genirq: generic chip: Add linear irq domain support

Quoted 11 lines · Russell King - ARM Linux

Review

Quoted 6 lines · Thomas Gleixner
(quoted diff) +6 -0
+	/* Init mask cache ? */
+	if (dgc->gc_flags & IRQ_GC_INIT_MASK_CACHE) {
+		raw_spin_lock_irqsave(&gc->lock, flags);
+		gc->mask_cache = irq_reg_readl(gc->reg_base + ct->regs.mask);
+		raw_spin_unlock_irqrestore(&gc->lock, flags);
+	}

This looks a little weird to me - it seems that it'll re-read this each time any irq is mapped in the domain, which is probably not wanted.

Yes, it's sloppy in two aspects.

1) It does not respect the per irq type mask cache, which got introduced in the same series

2) It rereads the mask cache for each mapping, but thats harmless because it's proper serialized. We can avoid that by clearing the IRQ_GC_INIT_MASK_CACHE bit when the first irq of that chip is mapped.

Congrats, you found a bug and as I said:

WARNING: It's compile tested only. So if you find bugs you can keep them and fix them yourself :)

Thanks,

tglx

author maintainer
#42 lore raw

Re: [RFC patch 7/8] genirq: generic chip: Add linear irq domain support

In reply to Thomas Gleixner
Quoted 2 lines · Thomas Gleixner

Provide infrastructure for irq chip implementations which work on linear irq domains.

Thomas,

I am happy that I put you into rant mode. It took me little more than an hour to read through your patches, prepare orion irqchip driver on top of them and finally got it working.

Anyway, I found some more issues.

Quoted 5 lines

Index: linux-2.6/kernel/irq/generic-chip.c
===================================================================

kernel/irq/generic-chip.c +0 -0
@@ -7,6 +7,7 @@

[...]

Quoted 1 line

+int irq_alloc_domain_generic_chips(struct irq_domain *d, int

irqs_per_chip,

Quoted 25 lines
kernel/irq/generic-chip.c +25 -0
+				   int num_ct, const char *name,
+				   irq_flow_handler_t handler,
+				   unsigned int clr, unsigned int set,
+				   enum irq_gc_flags gcflags)
+{
+	struct irq_domain_chip_generic *dgc;
+	struct irq_chip_generic *gc;
+	int numchips, sz, i;
+	unsigned long flags;
+
+	if (d->gc)
+		return -EBUSY;
+
+	if (d->revmap_type != IRQ_DOMAIN_MAP_LINEAR)
+		return -EINVAL;
+
+	numchips = d->revmap_data.linear.size / irqs_per_chip;
+	if (!numchips)
+		return -EINVAL;
+
+	sz = sizeof(*gc) + num_ct * sizeof(struct irq_chip_type);
+	sz *= numchips;
+	sz += sizeof(*dgc);
+
+	dgc = kzalloc(sz, GFP_KERNEL);

kernel/irq/generic-chip.c

Quoted 10 lines
kernel/irq/generic-chip.c +10 -0
+	if (!dgc)
+		return -ENOMEM;
+	dgc->irqs_per_chip = irqs_per_chip;
+	dgc->num_chips = numchips;
+	dgc->irq_flags_to_set = set;
+	dgc->irq_flags_to_clear = clr;
+	dgc->gc_flags = gcflags;
+	gc = dgc->gc;
+
+	for (i = 0; i < numchips; i++, gc++) {

The memory you allocated for gc, num_ct * ct, and dgc doesn't allow to increment through gc. gc is struct irq_chip_generic * but next gc is@sizeof(*gc) + num_ct * sizeof(struct irq_chip_type). This also affects indexing dgc->gc later.

I chose to fix it by having an index helper but that first maps dgc-gc to unsigned char * and then adds the correct offset. Not nice but it works. Maybe having real array of ptr to gc is more intuitive here even if we will have to have split kzallocs.

kernel/irq/generic-chip.c

Quoted 2 lines
kernel/irq/generic-chip.c +2 -0
+		irq_init_generic_chip(gc, name, num_ct, i * irqs_per_chip,
+				      NULL, handler);

irq_init_generic_chip does not take care of initalizing ct mask_cache ptr. This should be done here.

kernel/irq/generic-chip.c

Quoted 6 lines
kernel/irq/generic-chip.c +6 -0
+		gc->domain = d;
+		raw_spin_lock_irqsave(&gc_lock, flags);
+		list_add_tail(&gc->list, &gc_list);
+		raw_spin_unlock_irqrestore(&gc_lock, flags);
+	}
+	d->gc = dgc;

Moving this assignment above the for loop allows to get gc by index as indexing helper relies on domain, not domain generic chip.

You want me to prepare patches for the above? Maybe you can split your RFC into 1-6 and 7-8. Then you can have 1-6 applied independently of irq_domain_generic_chip stuff.

Thanks for the RFC again!

Sebastian

#43 lore raw

Re: [RFC patch 7/8] genirq: generic chip: Add linear irq domain support

Quoted 9 lines · Sebastian Hesselbarth
Quoted 2 lines · Thomas Gleixner

Provide infrastructure for irq chip implementations which work on linear irq domains.

Thomas,

I am happy that I put you into rant mode. It took me little more than an hour to read through your patches, prepare orion irqchip driver on top of them and finally got it working.

Cool.

Quoted 1 line

Anyway, I found some more issues.

That was expected. :)

Quoted 6 lines

Review

Quoted 1 line
(quoted diff) +1 -0
+	for (i = 0; i < numchips; i++, gc++) {

The memory you allocated for gc, num_ct * ct, and dgc doesn't allow to increment through gc. gc is struct irq_chip_generic * but next gc is at sizeof(*gc) + num_ct * sizeof(struct irq_chip_type). This also affects indexing dgc->gc later.

Indeed.

Quoted 2 lines

I chose to fix it by having an index helper but that first maps dgc-gc to unsigned char * and then adds the correct offset. Not

void * is the preferred over uchar *

Quoted 2 lines

nice but it works. Maybe having real array of ptr to gc is more intuitive here even if we will have to have split kzallocs.

No, you still can have a single kzalloc. It's just a matter of setting the pointers correctly.

Quoted 5 lines

Review

Quoted 2 lines
(quoted diff) +2 -0
+		irq_init_generic_chip(gc, name, num_ct, i * irqs_per_chip,
+				      NULL, handler);

irq_init_generic_chip does not take care of initalizing ct mask_cache ptr. This should be done here.

Right.

Quoted 11 lines

Review

Quoted 6 lines
(quoted diff) +6 -0
+		gc->domain = d;
+		raw_spin_lock_irqsave(&gc_lock, flags);
+		list_add_tail(&gc->list, &gc_list);
+		raw_spin_unlock_irqrestore(&gc_lock, flags);
+	}
+	d->gc = dgc;

Moving this assignment above the for loop allows to get gc by index as indexing helper relies on domain, not domain generic chip.

Quoted 1 line

You want me to prepare patches for the above? Maybe you can

That'd be nice.

Quoted 4 lines

split your RFC into 1-6 and 7-8. Then you can have 1-6 applied independently of irq_domain_generic_chip stuff.

Thanks for the RFC again!

Welcome. Have fun!

tglx

author maintainer
#44 lore raw

RFC PATCH 7/8 fixup 1/2: genirq: generic chip: Add linear irq domain support

In reply to Thomas Gleixner

irq_domain_chip_generic is allocating and indexing irq_chip_generic itself. However, the size of irq_chip_generic varies with number of irq_chip_types. This fixup moves irq_chip_generic helt by irq_domain_chip_generic to array of ptr and fixes the pointer arith used by irq_alloc_domain_generic_chip.

---
Cc: Thomas Gleixner <[email protected]>
Cc: Russell King - ARM Linux <[email protected]>
Cc: Grant Likely <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: Rob Landley <[email protected]>
Cc: Arnd Bergmann <[email protected]>
Cc: Jason Cooper <[email protected]>
Cc: Andrew Lunn <[email protected]>
Cc: Jason Gunthorpe <[email protected]>
Cc: Thomas Petazzoni <[email protected]>
Cc: Gregory Clement <[email protected]>
Cc: Ezequiel Garcia <[email protected]>
Cc: Maxime Ripard <[email protected]>
Cc: Jean-Francois Moine <[email protected]>
Cc: Gerlando Falauto <[email protected]>
Cc: Uwe Kleine-Koenig <[email protected]>
Cc: devicetree-discuss at lists.ozlabs.org
Cc: linux-doc at vger.kernel.org
Cc: linux-arm-kernel at lists.infradead.org
Cc: linux-kernel at vger.kernel.org

---
 include/linux/irq.h       |    4 ++--
 kernel/irq/generic-chip.c |   20 +++++++++++++-------
 2 files changed, 15 insertions(+), 9 deletions(-)
include/linux/irq.h +2 -2
index 7315155..fd2d7cb 100644
@@ -730,7 +730,7 @@ enum irq_gc_flags {
  * @irq_flags_to_set:	IRQ* flags to set on irq setup
  * @irq_flags_to_clear:	IRQ* flags to clear on irq setup
  * @gc_flags:		Generic chip specific setup flags
- * @gc:			Array of generic interrupt chips
+ * @gc:			Array of pointer to generic interrupt chips
  */
 struct irq_domain_chip_generic {
 	unsigned int		irqs_per_chip;
@@ -738,7 +738,7 @@ struct irq_domain_chip_generic {
 	unsigned int		irq_flags_to_clear;
 	unsigned int		irq_flags_to_set;
 	enum irq_gc_flags	gc_flags;
-	struct irq_chip_generic	gc[0];
+	struct irq_chip_generic	*gc[0];
 };
 
 /* Generic chip callback functions */
kernel/irq/generic-chip.c +13 -7
index e212b26..3dbfe2e 100644
@@ -247,6 +247,7 @@ int irq_alloc_domain_generic_chips(struct irq_domain *d, int irqs_per_chip,
 	struct irq_chip_generic *gc;
 	int numchips, sz, i;
 	unsigned long flags;
+	void *p;
 
 	if (d->gc)
 		return -EBUSY;
@@ -260,9 +261,9 @@ int irq_alloc_domain_generic_chips(struct irq_domain *d, int irqs_per_chip,
 
 	sz = sizeof(*gc) + num_ct * sizeof(struct irq_chip_type);
 	sz *= numchips;
-	sz += sizeof(*dgc);
+	sz += sizeof(*dgc) + numchips * sizeof(void *);
 
-	dgc = kzalloc(sz, GFP_KERNEL);
+	p = dgc = kzalloc(sz, GFP_KERNEL);
 	if (!dgc)
 		return -ENOMEM;
 	dgc->irqs_per_chip = irqs_per_chip;
@@ -270,17 +271,22 @@ int irq_alloc_domain_generic_chips(struct irq_domain *d, int irqs_per_chip,
 	dgc->irq_flags_to_set = set;
 	dgc->irq_flags_to_clear = clr;
 	dgc->gc_flags = gcflags;
-	gc = dgc->gc;
+	d->gc = dgc;
 
-	for (i = 0; i < numchips; i++, gc++) {
+	p += sizeof(*dgc) + numchips * sizeof(void *);
+	for (i = 0; i < numchips; i++) {
+		gc = (struct irq_chip_generic *)p;
+		dgc->gc[i] = gc;
 		irq_init_generic_chip(gc, name, num_ct, i * irqs_per_chip,
 				      NULL, handler);
 		gc->domain = d;
+
 		raw_spin_lock_irqsave(&gc_lock, flags);
 		list_add_tail(&gc->list, &gc_list);
 		raw_spin_unlock_irqrestore(&gc_lock, flags);
+
+		p += sizeof(*gc) + num_ct * sizeof(struct irq_chip_type);
 	}
-	d->gc = dgc;
 	return 0;
 }
 EXPORT_SYMBOL_GPL(irq_alloc_domain_generic_chips);
@@ -301,7 +307,7 @@ irq_get_domain_generic_chip(struct irq_domain *d, unsigned int hw_irq)
 	idx = hw_irq / dgc->irqs_per_chip;
 	if (idx >= dgc->num_chips)
 		return NULL;
-	return &dgc->gc[idx];
+	return dgc->gc[idx];
 }
 EXPORT_SYMBOL_GPL(irq_get_domain_generic_chip);
 
@@ -331,7 +337,7 @@ static int irq_map_generic_chip(struct irq_domain *d, unsigned int virq,
 	idx = hw_irq / dgc->irqs_per_chip;
 	if (idx >= dgc->num_chips)
 		return -EINVAL;
-	gc = &dgc->gc[idx];
+	gc = dgc->gc[idx];
 
 	idx = hw_irq % dgc->irqs_per_chip;
 
Signature · 1 line
1.7.2.5
author maintainer
#45 lore raw

RFC PATCH 7/8 fixup 2/2: genirq: generic chip: Add linear irq domain support

mask_cache pointer also needs to be initialized for domain generic chips.

---
Cc: Thomas Gleixner <[email protected]>
Cc: Russell King - ARM Linux <[email protected]>
Cc: Grant Likely <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: Rob Landley <[email protected]>
Cc: Arnd Bergmann <[email protected]>
Cc: Jason Cooper <[email protected]>
Cc: Andrew Lunn <[email protected]>
Cc: Jason Gunthorpe <[email protected]>
Cc: Thomas Petazzoni <[email protected]>
Cc: Gregory Clement <[email protected]>
Cc: Ezequiel Garcia <[email protected]>
Cc: Maxime Ripard <[email protected]>
Cc: Jean-Francois Moine <[email protected]>
Cc: Gerlando Falauto <[email protected]>
Cc: Uwe Kleine-Koenig <[email protected]>
Cc: devicetree-discuss at lists.ozlabs.org
Cc: linux-doc at vger.kernel.org
Cc: linux-arm-kernel at lists.infradead.org
Cc: linux-kernel at vger.kernel.org

---
 kernel/irq/generic-chip.c |   10 ++++++++++
 1 files changed, 10 insertions(+), 0 deletions(-)
kernel/irq/generic-chip.c +10 -0
index 3dbfe2e..3e0312f 100644
@@ -275,12 +275,22 @@ int irq_alloc_domain_generic_chips(struct irq_domain *d, int irqs_per_chip,
 
 	p += sizeof(*dgc) + numchips * sizeof(void *);
 	for (i = 0; i < numchips; i++) {
+		int k;
+
 		gc = (struct irq_chip_generic *)p;
 		dgc->gc[i] = gc;
 		irq_init_generic_chip(gc, name, num_ct, i * irqs_per_chip,
 				      NULL, handler);
 		gc->domain = d;
 
+		for (k = 0; k < gc->num_ct; k++) {
+			struct irq_chip_type *ct = &gc->chip_types[k];
+			if (gcflags & IRQ_GC_MASK_CACHE_PER_TYPE)
+				ct->mask_cache = &ct->mask_cache_priv;
+			else
+				ct->mask_cache = &gc->mask_cache;
+		}
+
 		raw_spin_lock_irqsave(&gc_lock, flags);
 		list_add_tail(&gc->list, &gc_list);
 		raw_spin_unlock_irqrestore(&gc_lock, flags);
Signature · 1 line
1.7.2.5
#46 lore raw

Re: [RFC patch 7/8] fixup 2/2: genirq: generic chip: Add linear irq domain support

Quoted 2 lines · Sebastian Hesselbarth

mask_cache pointer also needs to be initialized for domain generic chips.

It's not only the mask cache pointer. We also need to initialize the mask cache itself. I solved it by spitting out that code from irq_setup_generic_chip().

Thanks,

tglx

#47 lore raw

Re: [RFC patch 7/8] fixup 1/2: genirq: generic chip: Add linear irq domain support

In reply to Sebastian Hesselbarth
Quoted 5 lines · Sebastian Hesselbarth

irq_domain_chip_generic is allocating and indexing irq_chip_generic itself. However, the size of irq_chip_generic varies with number of irq_chip_types. This fixup moves irq_chip_generic helt by irq_domain_chip_generic to array of ptr and fixes the pointer arith used by irq_alloc_domain_generic_chip.

Thanks. I folded it in.

author maintainer
#49 lore raw

Re: [RFC patch 8/8] irqchip: sun4i: Convert to generic irq chip

Quoted 6 lines · Thomas Gleixner

Proof of concept patch to demonstrate the new irqdomain support for the generic irq chip. Untested !!

---

...

Review

Quoted 6 lines
(quoted diff) +6 -0
+		gc->reg_base = sun4i_irq_base;
+		gc->chip_types[0].regs.mask = SUN4I_IRQ_ENABLE_REG(i);
+		gc->chip_types[0].regs.ack = SUN4I_IRQ_PENDING_REG(i);
+		gc->chip_types[0].chip.mask = irq_gc_mask_clr_bit;
+		gc->chip_types[0].chip.ack = irq_gc_ack_set_bit;
+		gc->chip_types[0].chip.unmask = irq_gc_mask_set_bit;

To help Maxime debug, the three callbacks must be prepended by irq_, i.e. chip.irq_mask, ...

Sebastian

#50 lore raw

Re: [RFC patch 0/8] genirq: Support for irq domains in generic irq chip

In reply to Thomas Gleixner

Hello,

Quoted 17 lines · Thomas Gleixner

The ongoing device tree support for ARM is creating new irq chip drivers in drivers/irqchip/ in a frenzy. Quite some of them are ripping out the generic irq chip implementation from arch/arm/* and just creating the same mess of duplicated code again, which was cleaned up with the generic irq chip implementation with a lot of effort. Sigh!

I already prodded a few people in reviews to tackle that issue with no outcome. Even more sigh!

Poor Sebastian triggered me into rant mode, but he ad hoc volunteered to give it a try. YAY!

Though he asked for a bit of kickstart help. So I squeezed out a few spare cycles and implemented the basics as far as I think that they should work.

Quoted 8 lines

The following series contains the missing bits and pieces including a somehow forgotten and now slightly modified series from Gerlando adding support for irq chips which need separate mask caches for different chip (control flow) types.

At the moment this supports only linear irq domains, but it could be extended to other types as well if the need arises. Though the ARM chips are pretty much all about linear domains AFAICT.

Is there a tree/set of patches that have already fixed the issues pointed out by Russell and Sebastian? I'd like to use it to get forward with my nvic patch and want to avert double work and merging different approaches.

Best regards
Uwe

Signature · 2 lines
Pengutronix e.K.                           | Uwe Kleine-K?nig            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
#51 lore raw

cover genirq: Support for irq domains in generic irq chip - V2

In reply to Thomas Gleixner

Changes vs. V1:

- Fixed the generic chip pointer thinko (Sebastian Hesselbarth)

- Proper support for mask cache

- Read mask hardware only for the first map of an generic chip instance

- sun4i prefix irq functions proper

Thanks,

tglx

#53 lore raw

[tip:irq/core] genirq: Generic chip: Remove the local cur_regs() function

genirq: Generic chip: Remove the local cur_regs() function
Commit-ID:  cfeaa93f8a13ae9117ae20933a38a406de80849e
Gitweb:     http://git.kernel.org/tip/cfeaa93f8a13ae9117ae20933a38a406de80849e
Author:     Gerlando Falauto <[email protected]>
AuthorDate: Mon, 6 May 2013 14:30:17 +0000
Committer:  Thomas Gleixner <[email protected]>
CommitDate: Wed, 29 May 2013 10:57:09 +0200

genirq: Generic chip: Remove the local cur_regs() function

Since we already have an irq_data_get_chip_type() function which returns a pointer to irq_chip_type, use that instead of cur_regs().

Cc Andrew Lunn <[email protected]>
Cc Joey Oravec <[email protected]>
Cc Lennert Buytenhek <[email protected]>
Cc Russell King - ARM Linux <[email protected]>
Cc Jason Gunthorpe <[email protected]>
Cc Holger Brunck <[email protected]>
Cc Ezequiel Garcia <[email protected]>
Cc Sebastian Hesselbarth <[email protected]>
Cc Jason Cooper <[email protected]>
Cc Arnd Bergmann <[email protected]>
Cc Rob Herring <[email protected]>
Cc Ben Dooks <[email protected]>
Cc Gregory Clement <[email protected]>
Cc Simon Guinot <[email protected]>
Cc Thomas Petazzoni <[email protected]>
Cc Jean-Francois Moine <[email protected]>
Cc Nicolas Pitre <[email protected]>
Cc Rob Landley <[email protected]>
Cc Maxime Ripard <[email protected]>
---
 kernel/irq/generic-chip.c | 31 +++++++++++++++++--------------
 1 file changed, 17 insertions(+), 14 deletions(-)
kernel/irq/generic-chip.c +17 -14
index c89295a..0e6ba78 100644
@@ -16,11 +16,6 @@
 static LIST_HEAD(gc_list);
 static DEFINE_RAW_SPINLOCK(gc_lock);
 
-static inline struct irq_chip_regs *cur_regs(struct irq_data *d)
-{
-	return &container_of(d->chip, struct irq_chip_type, chip)->regs;
-}
-
 /**
  * irq_gc_noop - NOOP function
  * @d: irq_data
@@ -39,10 +34,11 @@ void irq_gc_noop(struct irq_data *d)
 void irq_gc_mask_disable_reg(struct irq_data *d)
 {
 	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
+	struct irq_chip_type *ct = irq_data_get_chip_type(d);
 	u32 mask = 1 << (d->irq - gc->irq_base);
 
 	irq_gc_lock(gc);
-	irq_reg_writel(mask, gc->reg_base + cur_regs(d)->disable);
+	irq_reg_writel(mask, gc->reg_base + ct->regs.disable);
 	gc->mask_cache &= ~mask;
 	irq_gc_unlock(gc);
 }
@@ -57,11 +53,12 @@ void irq_gc_mask_disable_reg(struct irq_data *d)
 void irq_gc_mask_set_bit(struct irq_data *d)
 {
 	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
+	struct irq_chip_type *ct = irq_data_get_chip_type(d);
 	u32 mask = 1 << (d->irq - gc->irq_base);
 
 	irq_gc_lock(gc);
 	gc->mask_cache |= mask;
-	irq_reg_writel(gc->mask_cache, gc->reg_base + cur_regs(d)->mask);
+	irq_reg_writel(gc->mask_cache, gc->reg_base + ct->regs.mask);
 	irq_gc_unlock(gc);
 }
 
@@ -75,11 +72,12 @@ void irq_gc_mask_set_bit(struct irq_data *d)
 void irq_gc_mask_clr_bit(struct irq_data *d)
 {
 	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
+	struct irq_chip_type *ct = irq_data_get_chip_type(d);
 	u32 mask = 1 << (d->irq - gc->irq_base);
 
 	irq_gc_lock(gc);
 	gc->mask_cache &= ~mask;
-	irq_reg_writel(gc->mask_cache, gc->reg_base + cur_regs(d)->mask);
+	irq_reg_writel(gc->mask_cache, gc->reg_base + ct->regs.mask);
 	irq_gc_unlock(gc);
 }
 
@@ -93,10 +91,11 @@ void irq_gc_mask_clr_bit(struct irq_data *d)
 void irq_gc_unmask_enable_reg(struct irq_data *d)
 {
 	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
+	struct irq_chip_type *ct = irq_data_get_chip_type(d);
 	u32 mask = 1 << (d->irq - gc->irq_base);
 
 	irq_gc_lock(gc);
-	irq_reg_writel(mask, gc->reg_base + cur_regs(d)->enable);
+	irq_reg_writel(mask, gc->reg_base + ct->regs.enable);
 	gc->mask_cache |= mask;
 	irq_gc_unlock(gc);
 }
@@ -108,10 +107,11 @@ void irq_gc_unmask_enable_reg(struct irq_data *d)
 void irq_gc_ack_set_bit(struct irq_data *d)
 {
 	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
+	struct irq_chip_type *ct = irq_data_get_chip_type(d);
 	u32 mask = 1 << (d->irq - gc->irq_base);
 
 	irq_gc_lock(gc);
-	irq_reg_writel(mask, gc->reg_base + cur_regs(d)->ack);
+	irq_reg_writel(mask, gc->reg_base + ct->regs.ack);
 	irq_gc_unlock(gc);
 }
 
@@ -122,10 +122,11 @@ void irq_gc_ack_set_bit(struct irq_data *d)
 void irq_gc_ack_clr_bit(struct irq_data *d)
 {
 	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
+	struct irq_chip_type *ct = irq_data_get_chip_type(d);
 	u32 mask = ~(1 << (d->irq - gc->irq_base));
 
 	irq_gc_lock(gc);
-	irq_reg_writel(mask, gc->reg_base + cur_regs(d)->ack);
+	irq_reg_writel(mask, gc->reg_base + ct->regs.ack);
 	irq_gc_unlock(gc);
 }
 
@@ -136,11 +137,12 @@ void irq_gc_ack_clr_bit(struct irq_data *d)
 void irq_gc_mask_disable_reg_and_ack(struct irq_data *d)
 {
 	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
+	struct irq_chip_type *ct = irq_data_get_chip_type(d);
 	u32 mask = 1 << (d->irq - gc->irq_base);
 
 	irq_gc_lock(gc);
-	irq_reg_writel(mask, gc->reg_base + cur_regs(d)->mask);
-	irq_reg_writel(mask, gc->reg_base + cur_regs(d)->ack);
+	irq_reg_writel(mask, gc->reg_base + ct->regs.mask);
+	irq_reg_writel(mask, gc->reg_base + ct->regs.ack);
 	irq_gc_unlock(gc);
 }
 
@@ -151,10 +153,11 @@ void irq_gc_mask_disable_reg_and_ack(struct irq_data *d)
 void irq_gc_eoi(struct irq_data *d)
 {
 	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
+	struct irq_chip_type *ct = irq_data_get_chip_type(d);
 	u32 mask = 1 << (d->irq - gc->irq_base);
 
 	irq_gc_lock(gc);
-	irq_reg_writel(mask, gc->reg_base + cur_regs(d)->eoi);
+	irq_reg_writel(mask, gc->reg_base + ct->regs.eoi);
 	irq_gc_unlock(gc);
 }
 
#55 lore raw

[tip:irq/core] genirq: Generic chip: Add support for per chip type mask cache

genirq: Generic chip: Add support for per chip type mask cache
Commit-ID:  899f0e66fff36ebb6dd6a83af9aa631f6cb7e0dc
Gitweb:     http://git.kernel.org/tip/899f0e66fff36ebb6dd6a83af9aa631f6cb7e0dc
Author:     Gerlando Falauto <[email protected]>
AuthorDate: Mon, 6 May 2013 14:30:19 +0000
Committer:  Thomas Gleixner <[email protected]>
CommitDate: Wed, 29 May 2013 10:57:10 +0200

genirq: Generic chip: Add support for per chip type mask cache

Today the same interrupt mask cache (stored within struct irq_chip_generic) is shared between all the irq_chip_type instances. As there are instances where each irq_chip_type uses a distinct mask register (as it is the case for Orion SoCs), sharing a single mask cache may be incorrect. So add a distinct pointer for each irq_chip_type, which for now points to the original mask register within irq_chip_generic. So no functional changes here.

[ tglx: Minor cosmetic tweaks ]

Cc Andrew Lunn <[email protected]>
Cc Lennert Buytenhek <[email protected]>
Cc Russell King - ARM Linux <[email protected]>
Cc Jason Gunthorpe <[email protected]>
Cc Holger Brunck <[email protected]>
Cc Ezequiel Garcia <[email protected]>
Cc Sebastian Hesselbarth <[email protected]>
Cc Jason Cooper <[email protected]>
Cc Arnd Bergmann <[email protected]>
Cc Rob Herring <[email protected]>
Cc Ben Dooks <[email protected]>
Cc Gregory Clement <[email protected]>
Cc Simon Guinot <[email protected]>
Cc Thomas Petazzoni <[email protected]>
Cc Jean-Francois Moine <[email protected]>
Cc Nicolas Pitre <[email protected]>
Cc Rob Landley <[email protected]>
Cc Maxime Ripard <[email protected]>
---
 include/linux/irq.h       |  6 +++++-
 kernel/irq/generic-chip.c | 16 ++++++++++------
 2 files changed, 15 insertions(+), 7 deletions(-)
include/linux/irq.h +5 -1
index bc4e066..38709a3 100644
@@ -644,6 +644,8 @@ struct irq_chip_regs {
  * @regs:		Register offsets for this chip
  * @handler:		Flow handler associated with this chip
  * @type:		Chip can handle these flow types
+ * @mask_cache_priv:	Cached mask register private to the chip type
+ * @mask_cache:		Pointer to cached mask register
  *
  * A irq_generic_chip can have several instances of irq_chip_type when
  * it requires different functions and register offsets for different
@@ -654,6 +656,8 @@ struct irq_chip_type {
 	struct irq_chip_regs	regs;
 	irq_flow_handler_t	handler;
 	u32			type;
+	u32			mask_cache_priv;
+	u32			*mask_cache;
 };
 
 /**
@@ -662,7 +666,7 @@ struct irq_chip_type {
  * @reg_base:		Register base address (virtual)
  * @irq_base:		Interrupt base nr for this chip
  * @irq_cnt:		Number of interrupts handled by this chip
- * @mask_cache:		Cached mask register
+ * @mask_cache:		Cached mask register shared between all chip types
  * @type_cache:		Cached type register
  * @polarity_cache:	Cached polarity register
  * @wake_enabled:	Interrupt can wakeup from suspend
kernel/irq/generic-chip.c +10 -6
index 0e6ba78..113d9eb 100644
@@ -39,7 +39,7 @@ void irq_gc_mask_disable_reg(struct irq_data *d)
 
 	irq_gc_lock(gc);
 	irq_reg_writel(mask, gc->reg_base + ct->regs.disable);
-	gc->mask_cache &= ~mask;
+	*ct->mask_cache &= ~mask;
 	irq_gc_unlock(gc);
 }
 
@@ -57,8 +57,8 @@ void irq_gc_mask_set_bit(struct irq_data *d)
 	u32 mask = 1 << (d->irq - gc->irq_base);
 
 	irq_gc_lock(gc);
-	gc->mask_cache |= mask;
-	irq_reg_writel(gc->mask_cache, gc->reg_base + ct->regs.mask);
+	*ct->mask_cache |= mask;
+	irq_reg_writel(*ct->mask_cache, gc->reg_base + ct->regs.mask);
 	irq_gc_unlock(gc);
 }
 
@@ -76,8 +76,8 @@ void irq_gc_mask_clr_bit(struct irq_data *d)
 	u32 mask = 1 << (d->irq - gc->irq_base);
 
 	irq_gc_lock(gc);
-	gc->mask_cache &= ~mask;
-	irq_reg_writel(gc->mask_cache, gc->reg_base + ct->regs.mask);
+	*ct->mask_cache &= ~mask;
+	irq_reg_writel(*ct->mask_cache, gc->reg_base + ct->regs.mask);
 	irq_gc_unlock(gc);
 }
 
@@ -96,7 +96,7 @@ void irq_gc_unmask_enable_reg(struct irq_data *d)
 
 	irq_gc_lock(gc);
 	irq_reg_writel(mask, gc->reg_base + ct->regs.enable);
-	gc->mask_cache |= mask;
+	*ct->mask_cache |= mask;
 	irq_gc_unlock(gc);
 }
 
@@ -250,6 +250,10 @@ void irq_setup_generic_chip(struct irq_chip_generic *gc, u32 msk,
 	if (flags & IRQ_GC_INIT_MASK_CACHE)
 		gc->mask_cache = irq_reg_readl(gc->reg_base + ct->regs.mask);
 
+	/* Initialize mask cache pointer */
+	for (i = 0; i < gc->num_ct; i++)
+		ct[i].mask_cache = &gc->mask_cache;
+
 	for (i = gc->irq_base; msk; msk >>= 1, i++) {
 		if (!(msk & 0x01))
 			continue;
#57 lore raw

[tip:irq/core] genirq: Generic chip: Handle separate mask registers

genirq: Generic chip: Handle separate mask registers
Commit-ID:  af80b0fed67261dcba2ce2406db1d553d07cbe75
Gitweb:     http://git.kernel.org/tip/af80b0fed67261dcba2ce2406db1d553d07cbe75
Author:     Gerlando Falauto <[email protected]>
AuthorDate: Mon, 6 May 2013 14:30:21 +0000
Committer:  Thomas Gleixner <[email protected]>
CommitDate: Wed, 29 May 2013 10:57:10 +0200

genirq: Generic chip: Handle separate mask registers

There are cases where all irq_chip_type instances have separate mask registers, making a shared mask register cache unsuitable for the purpose.

Introduce a new flag IRQ_GC_MASK_CACHE_PER_TYPE. If set, point the per chip mask pointer to the per chip private mask cache instead.

[ tglx: Simplified code, renamed flag and massaged changelog ]

Cc Andrew Lunn <[email protected]>
Cc Joey Oravec <[email protected]>
Cc Lennert Buytenhek <[email protected]>
Cc Russell King - ARM Linux <[email protected]>
Cc Jason Gunthorpe <[email protected]>
Cc Holger Brunck <[email protected]>
Cc Ezequiel Garcia <[email protected]>
Cc Sebastian Hesselbarth <[email protected]>
Cc Jason Cooper <[email protected]>
Cc Arnd Bergmann <[email protected]>
Cc Rob Herring <[email protected]>
Cc Ben Dooks <[email protected]>
Cc Gregory Clement <[email protected]>
Cc Simon Guinot <[email protected]>
Cc Thomas Petazzoni <[email protected]>
Cc Jean-Francois Moine <[email protected]>
Cc Nicolas Pitre <[email protected]>
Cc Rob Landley <[email protected]>
Cc Maxime Ripard <[email protected]>
---
 include/linux/irq.h       |  2 ++
 kernel/irq/generic-chip.c | 17 ++++++++++-------
 2 files changed, 12 insertions(+), 7 deletions(-)
include/linux/irq.h +2 -0
index 38709a3..7f1f015 100644
@@ -704,10 +704,12 @@ struct irq_chip_generic {
  * @IRQ_GC_INIT_NESTED_LOCK:	Set the lock class of the irqs to nested for
  *				irq chips which need to call irq_set_wake() on
  *				the parent irq. Usually GPIO implementations
+ * @IRQ_GC_MASK_CACHE_PER_TYPE:	Mask cache is chip type private
  */
 enum irq_gc_flags {
 	IRQ_GC_INIT_MASK_CACHE		= 1 << 0,
 	IRQ_GC_INIT_NESTED_LOCK		= 1 << 1,
+	IRQ_GC_MASK_CACHE_PER_TYPE	= 1 << 2,
 };
 
 /* Generic chip callback functions */
kernel/irq/generic-chip.c +10 -7
index 113d9eb..da2a941 100644
@@ -241,18 +241,21 @@ void irq_setup_generic_chip(struct irq_chip_generic *gc, u32 msk,
 {
 	struct irq_chip_type *ct = gc->chip_types;
 	unsigned int i;
+	u32 *mskptr = &gc->mask_cache, mskreg = ct->regs.mask;
 
 	raw_spin_lock(&gc_lock);
 	list_add_tail(&gc->list, &gc_list);
 	raw_spin_unlock(&gc_lock);
 
-	/* Init mask cache ? */
-	if (flags & IRQ_GC_INIT_MASK_CACHE)
-		gc->mask_cache = irq_reg_readl(gc->reg_base + ct->regs.mask);
-
-	/* Initialize mask cache pointer */
-	for (i = 0; i < gc->num_ct; i++)
-		ct[i].mask_cache = &gc->mask_cache;
+	for (i = 0; i < gc->num_ct; i++) {
+		if (flags & IRQ_GC_MASK_CACHE_PER_TYPE) {
+			mskptr = &ct[i].mask_cache_priv;
+			mskreg = ct[i].regs.mask;
+		}
+		ct[i].mask_cache = mskptr;
+		if (flags & IRQ_GC_INIT_MASK_CACHE)
+			*mskptr = irq_reg_readl(gc->reg_base + mskreg);
+	}
 
 	for (i = gc->irq_base; msk; msk >>= 1, i++) {
 		if (!(msk & 0x01))
#59 lore raw

[tip:irq/core] genirq: Generic chip: Cache per irq bit mask

genirq: Generic chip: Cache per irq bit mask
Commit-ID:  966dc736b819999cd2d3a6408d47d33b579f7d56
Gitweb:     http://git.kernel.org/tip/966dc736b819999cd2d3a6408d47d33b579f7d56
Author:     Thomas Gleixner <[email protected]>
AuthorDate: Mon, 6 May 2013 14:30:22 +0000
Committer:  Thomas Gleixner <[email protected]>
CommitDate: Wed, 29 May 2013 10:57:10 +0200

genirq: Generic chip: Cache per irq bit mask

Cache the per irq bit mask instead of recalculating it over and over.

Cc Thomas Petazzoni <[email protected]>
Cc Andrew Lunn <[email protected]>
Cc Russell King - ARM Linux <[email protected]>
Cc Jason Cooper <[email protected]>
Cc Arnd Bergmann <[email protected]>
Cc Jean-Francois Moine <[email protected]>
Cc Rob Herring <[email protected]>
Cc Jason Gunthorpe <[email protected]>
Cc Gregory Clement <[email protected]>
Cc Gerlando Falauto <[email protected]>
Cc Rob Landley <[email protected]>
Cc Maxime Ripard <[email protected]>
Cc Ezequiel Garcia <[email protected]>
Cc Sebastian Hesselbarth <[email protected]>
---
 include/linux/irq.h       |  4 ++++
 kernel/irq/generic-chip.c | 23 ++++++++++++++---------
 2 files changed, 18 insertions(+), 9 deletions(-)
include/linux/irq.h +4 -0
index 7f1f015..d5fc7f5 100644
@@ -119,6 +119,7 @@ struct irq_domain;
 
 /**
  * struct irq_data - per irq and irq chip data passed down to chip functions
+ * @mask:		precomputed bitmask for accessing the chip registers
  * @irq:		interrupt number
  * @hwirq:		hardware interrupt number, local to the interrupt domain
  * @node:		node index useful for balancing
@@ -138,6 +139,7 @@ struct irq_domain;
  * irq_data.
  */
 struct irq_data {
+	u32			mask;
 	unsigned int		irq;
 	unsigned long		hwirq;
 	unsigned int		node;
@@ -705,11 +707,13 @@ struct irq_chip_generic {
  *				irq chips which need to call irq_set_wake() on
  *				the parent irq. Usually GPIO implementations
  * @IRQ_GC_MASK_CACHE_PER_TYPE:	Mask cache is chip type private
+ * @IRQ_GC_NO_MASK:		Do not calculate irq_data->mask
  */
 enum irq_gc_flags {
 	IRQ_GC_INIT_MASK_CACHE		= 1 << 0,
 	IRQ_GC_INIT_NESTED_LOCK		= 1 << 1,
 	IRQ_GC_MASK_CACHE_PER_TYPE	= 1 << 2,
+	IRQ_GC_NO_MASK			= 1 << 3,
 };
 
 /* Generic chip callback functions */
kernel/irq/generic-chip.c +14 -9
index da2a941..957155c 100644
@@ -35,7 +35,7 @@ void irq_gc_mask_disable_reg(struct irq_data *d)
 {
 	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
 	struct irq_chip_type *ct = irq_data_get_chip_type(d);
-	u32 mask = 1 << (d->irq - gc->irq_base);
+	u32 mask = d->mask;
 
 	irq_gc_lock(gc);
 	irq_reg_writel(mask, gc->reg_base + ct->regs.disable);
@@ -54,7 +54,7 @@ void irq_gc_mask_set_bit(struct irq_data *d)
 {
 	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
 	struct irq_chip_type *ct = irq_data_get_chip_type(d);
-	u32 mask = 1 << (d->irq - gc->irq_base);
+	u32 mask = d->mask;
 
 	irq_gc_lock(gc);
 	*ct->mask_cache |= mask;
@@ -73,7 +73,7 @@ void irq_gc_mask_clr_bit(struct irq_data *d)
 {
 	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
 	struct irq_chip_type *ct = irq_data_get_chip_type(d);
-	u32 mask = 1 << (d->irq - gc->irq_base);
+	u32 mask = d->mask;
 
 	irq_gc_lock(gc);
 	*ct->mask_cache &= ~mask;
@@ -92,7 +92,7 @@ void irq_gc_unmask_enable_reg(struct irq_data *d)
 {
 	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
 	struct irq_chip_type *ct = irq_data_get_chip_type(d);
-	u32 mask = 1 << (d->irq - gc->irq_base);
+	u32 mask = d->mask;
 
 	irq_gc_lock(gc);
 	irq_reg_writel(mask, gc->reg_base + ct->regs.enable);
@@ -108,7 +108,7 @@ void irq_gc_ack_set_bit(struct irq_data *d)
 {
 	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
 	struct irq_chip_type *ct = irq_data_get_chip_type(d);
-	u32 mask = 1 << (d->irq - gc->irq_base);
+	u32 mask = d->mask;
 
 	irq_gc_lock(gc);
 	irq_reg_writel(mask, gc->reg_base + ct->regs.ack);
@@ -123,7 +123,7 @@ void irq_gc_ack_clr_bit(struct irq_data *d)
 {
 	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
 	struct irq_chip_type *ct = irq_data_get_chip_type(d);
-	u32 mask = ~(1 << (d->irq - gc->irq_base));
+	u32 mask = ~d->mask;
 
 	irq_gc_lock(gc);
 	irq_reg_writel(mask, gc->reg_base + ct->regs.ack);
@@ -138,7 +138,7 @@ void irq_gc_mask_disable_reg_and_ack(struct irq_data *d)
 {
 	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
 	struct irq_chip_type *ct = irq_data_get_chip_type(d);
-	u32 mask = 1 << (d->irq - gc->irq_base);
+	u32 mask = d->mask;
 
 	irq_gc_lock(gc);
 	irq_reg_writel(mask, gc->reg_base + ct->regs.mask);
@@ -154,7 +154,7 @@ void irq_gc_eoi(struct irq_data *d)
 {
 	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
 	struct irq_chip_type *ct = irq_data_get_chip_type(d);
-	u32 mask = 1 << (d->irq - gc->irq_base);
+	u32 mask = d->mask;
 
 	irq_gc_lock(gc);
 	irq_reg_writel(mask, gc->reg_base + ct->regs.eoi);
@@ -172,7 +172,7 @@ void irq_gc_eoi(struct irq_data *d)
 int irq_gc_set_wake(struct irq_data *d, unsigned int on)
 {
 	struct irq_chip_generic *gc = irq_data_get_irq_chip_data(d);
-	u32 mask = 1 << (d->irq - gc->irq_base);
+	u32 mask = d->mask;
 
 	if (!(mask & gc->wake_enabled))
 		return -EINVAL;
@@ -264,6 +264,11 @@ void irq_setup_generic_chip(struct irq_chip_generic *gc, u32 msk,
 		if (flags & IRQ_GC_INIT_NESTED_LOCK)
 			irq_set_lockdep_class(i, &irq_nested_lock_class);
 
+		if (!(flags & IRQ_GC_NO_MASK)) {
+			struct irq_data *d = irq_get_irq_data(i);
+
+			d->mask = 1 << (i - gc->irq_base);
+		}
 		irq_set_chip_and_handler(i, &ct->chip, ct->handler);
 		irq_set_chip_data(i, gc);
 		irq_modify_status(i, clr, set);
#61 lore raw

[tip:irq/core] genirq: irqchip: Add a mask calculation function

genirq: irqchip: Add a mask calculation function
Commit-ID:  d0051816e619f8f082582bec07ffa51bdb4f2104
Gitweb:     http://git.kernel.org/tip/d0051816e619f8f082582bec07ffa51bdb4f2104
Author:     Thomas Gleixner <[email protected]>
AuthorDate: Mon, 6 May 2013 14:30:24 +0000
Committer:  Thomas Gleixner <[email protected]>
CommitDate: Wed, 29 May 2013 10:57:10 +0200

genirq: irqchip: Add a mask calculation function

Some chips have weird bit mask access patterns instead of the linear you expect. Allow them to calculate the cached mask themself.

Cc Thomas Petazzoni <[email protected]>
Cc Andrew Lunn <[email protected]>
Cc Russell King - ARM Linux <[email protected]>
Cc Jason Cooper <[email protected]>
Cc Arnd Bergmann <[email protected]>
Cc Jean-Francois Moine <[email protected]>
Cc Rob Herring <[email protected]>
Cc Jason Gunthorpe <[email protected]>
Cc Gregory Clement <[email protected]>
Cc Gerlando Falauto <[email protected]>
Cc Rob Landley <[email protected]>
Cc Maxime Ripard <[email protected]>
Cc Ezequiel Garcia <[email protected]>
Cc Sebastian Hesselbarth <[email protected]>
---
 include/linux/irq.h       | 3 +++
 kernel/irq/generic-chip.c | 8 ++++++--
 2 files changed, 9 insertions(+), 2 deletions(-)
include/linux/irq.h +3 -0
index d5fc7f5..ab8169f 100644
@@ -296,6 +296,7 @@ static inline irq_hw_number_t irqd_to_hwirq(struct irq_data *d)
  * @irq_suspend:	function called from core code on suspend once per chip
  * @irq_resume:		function called from core code on resume once per chip
  * @irq_pm_shutdown:	function called from core code on shutdown once per chip
+ * @irq_calc_mask:	Optional function to set irq_data.mask for special cases
  * @irq_print_chip:	optional to print special chip info in show_interrupts
  * @flags:		chip specific flags
  */
@@ -327,6 +328,8 @@ struct irq_chip {
 	void		(*irq_resume)(struct irq_data *data);
 	void		(*irq_pm_shutdown)(struct irq_data *data);
 
+	void		(*irq_calc_mask)(struct irq_data *data);
+
 	void		(*irq_print_chip)(struct irq_data *data, struct seq_file *p);
 
 	unsigned long	flags;
kernel/irq/generic-chip.c +6 -2
index 957155c..5068fe3 100644
@@ -240,6 +240,7 @@ void irq_setup_generic_chip(struct irq_chip_generic *gc, u32 msk,
 			    unsigned int set)
 {
 	struct irq_chip_type *ct = gc->chip_types;
+	struct irq_chip *chip = &ct->chip;
 	unsigned int i;
 	u32 *mskptr = &gc->mask_cache, mskreg = ct->regs.mask;
 
@@ -267,9 +268,12 @@ void irq_setup_generic_chip(struct irq_chip_generic *gc, u32 msk,
 		if (!(flags & IRQ_GC_NO_MASK)) {
 			struct irq_data *d = irq_get_irq_data(i);
 
-			d->mask = 1 << (i - gc->irq_base);
+			if (chip->irq_calc_mask)
+				chip->irq_calc_mask(d);
+			else
+				d->mask = 1 << (i - gc->irq_base);
 		}
-		irq_set_chip_and_handler(i, &ct->chip, ct->handler);
+		irq_set_chip_and_handler(i, chip, ct->handler);
 		irq_set_chip_data(i, gc);
 		irq_modify_status(i, clr, set);
 	}
#63 lore raw

[tip:irq/core] genirq: Generic chip: Split out code into separate functions

genirq: Generic chip: Split out code into separate functions
Commit-ID:  3528d82b684680b72fa31881c8c572c5a98b51de
Gitweb:     http://git.kernel.org/tip/3528d82b684680b72fa31881c8c572c5a98b51de
Author:     Thomas Gleixner <[email protected]>
AuthorDate: Mon, 6 May 2013 14:30:25 +0000
Committer:  Thomas Gleixner <[email protected]>
CommitDate: Wed, 29 May 2013 10:57:11 +0200

genirq: Generic chip: Split out code into separate functions

Preparatory patch for linear interrupt domains.

Cc Thomas Petazzoni <[email protected]>
Cc Andrew Lunn <[email protected]>
Cc Russell King - ARM Linux <[email protected]>
Cc Jason Cooper <[email protected]>
Cc Arnd Bergmann <[email protected]>
Cc Jean-Francois Moine <[email protected]>
Cc Rob Herring <[email protected]>
Cc Jason Gunthorpe <[email protected]>
Cc Gregory Clement <[email protected]>
Cc Gerlando Falauto <[email protected]>
Cc Rob Landley <[email protected]>
Cc Maxime Ripard <[email protected]>
Cc Ezequiel Garcia <[email protected]>
Cc Sebastian Hesselbarth <[email protected]>
---
 kernel/irq/generic-chip.c | 50 ++++++++++++++++++++++++++++++++---------------
 1 file changed, 34 insertions(+), 16 deletions(-)
kernel/irq/generic-chip.c +34 -16
index 5068fe3..3deb333 100644
@@ -186,6 +186,19 @@ int irq_gc_set_wake(struct irq_data *d, unsigned int on)
 	return 0;
 }
 
+static void
+irq_init_generic_chip(struct irq_chip_generic *gc, const char *name,
+		      int num_ct, unsigned int irq_base,
+		      void __iomem *reg_base, irq_flow_handler_t handler)
+{
+	raw_spin_lock_init(&gc->lock);
+	gc->num_ct = num_ct;
+	gc->irq_base = irq_base;
+	gc->reg_base = reg_base;
+	gc->chip_types->chip.name = name;
+	gc->chip_types->handler = handler;
+}
+
 /**
  * irq_alloc_generic_chip - Allocate a generic chip and initialize it
  * @name:	Name of the irq chip
@@ -206,17 +219,31 @@ irq_alloc_generic_chip(const char *name, int num_ct, unsigned int irq_base,
 
 	gc = kzalloc(sz, GFP_KERNEL);
 	if (gc) {
-		raw_spin_lock_init(&gc->lock);
-		gc->num_ct = num_ct;
-		gc->irq_base = irq_base;
-		gc->reg_base = reg_base;
-		gc->chip_types->chip.name = name;
-		gc->chip_types->handler = handler;
+		irq_init_generic_chip(gc, name, num_ct, irq_base, reg_base,
+				      handler);
 	}
 	return gc;
 }
 EXPORT_SYMBOL_GPL(irq_alloc_generic_chip);
 
+static void
+irq_gc_init_mask_cache(struct irq_chip_generic *gc, enum irq_gc_flags flags)
+{
+	struct irq_chip_type *ct = gc->chip_types;
+	u32 *mskptr = &gc->mask_cache, mskreg = ct->regs.mask;
+	int i;
+
+	for (i = 0; i < gc->num_ct; i++) {
+		if (flags & IRQ_GC_MASK_CACHE_PER_TYPE) {
+			mskptr = &ct[i].mask_cache_priv;
+			mskreg = ct[i].regs.mask;
+		}
+		ct[i].mask_cache = mskptr;
+		if (flags & IRQ_GC_INIT_MASK_CACHE)
+			*mskptr = irq_reg_readl(gc->reg_base + mskreg);
+	}
+}
+
 /*
  * Separate lockdep class for interrupt chip which can nest irq_desc
  * lock.
@@ -242,21 +269,12 @@ void irq_setup_generic_chip(struct irq_chip_generic *gc, u32 msk,
 	struct irq_chip_type *ct = gc->chip_types;
 	struct irq_chip *chip = &ct->chip;
 	unsigned int i;
-	u32 *mskptr = &gc->mask_cache, mskreg = ct->regs.mask;
 
 	raw_spin_lock(&gc_lock);
 	list_add_tail(&gc->list, &gc_list);
 	raw_spin_unlock(&gc_lock);
 
-	for (i = 0; i < gc->num_ct; i++) {
-		if (flags & IRQ_GC_MASK_CACHE_PER_TYPE) {
-			mskptr = &ct[i].mask_cache_priv;
-			mskreg = ct[i].regs.mask;
-		}
-		ct[i].mask_cache = mskptr;
-		if (flags & IRQ_GC_INIT_MASK_CACHE)
-			*mskptr = irq_reg_readl(gc->reg_base + mskreg);
-	}
+	irq_gc_init_mask_cache(gc, flags);
 
 	for (i = gc->irq_base; msk; msk >>= 1, i++) {
 		if (!(msk & 0x01))
Acked-by on v1 7/8
#65 lore raw

Re: [patch 7/8] genirq: generic chip: Add linear irq domain support

Quoted 12 lines · Thomas Gleixner

Provide infrastructure for irq chip implementations which work on linear irq domains.

- Interface to allocate multiple generic chips which are associated to the irq domain.

- Interface to get the generic chip pointer for a particular hardware interrupt in the domain.

- irq domain mapping function to install the chip for a particular interrupt.

Quoted 6 lines

Note: This lacks a removal function for now, but this is a draft patch the ARM folks to work on.

[ Sebastian Hesselbarth: Mask cache and pointer math fixups ]

Hi Thomas,

But I have some comments below...

Quoted 21 lines
---
 include/linux/irq.h       |   30 +++++++
 include/linux/irqdomain.h |   12 ++
 kernel/irq/generic-chip.c |  187 ++++++++++++++++++++++++++++++++++++++++++++--
 kernel/irq/irqdomain.c    |    6 -
 4 files changed, 223 insertions(+), 12 deletions(-)

Index: linux-2.6/include/linux/irq.h
===================================================================

include/linux/irq.h +2 -0
@@ -678,6 +678,8 @@ struct irq_chip_type {
  * @wake_active:	Interrupt is marked as an wakeup from suspend source
  * @num_ct:		Number of available irq_chip_type instances (usually 1)
  * @private:		Private data for non generic chip callbacks
+ * @installed:		bitfield to denote installed interrupts
+ * @domain:		irq domain pointer
  * @list:		List head for keeping track of instances
  * @chip_types:		Array of interrupt irq_chip_types
  *

include/linux/irq.h struct irq_chip_generic { · line 704

Quoted 4 lines
include/linux/irq.h +1 -0
@@ -699,6 +701,8 @@ struct irq_chip_generic {
 	u32			wake_active;
 	unsigned int		num_ct;
 	void			*private;
+	unsigned long		installed;

This is probably something that the irqdomain should be keeping track of internally, but that's an issue for a separate patch series.

[...]

include/linux/irq.h struct irq_chip_generic { · line 707

Quoted 3 lines
include/linux/irq.h +3 -0
+struct irq_domain_ops irq_generic_chip_ops = {
+	.map = irq_map_generic_chip,
+	.xlate = irq_domain_xlate_onecell,

As discussed on IRC, should use onetwocell here for greater compatibility with existing bindings.

Cheers,
g.

#66 lore raw

Re: [patch 7/8] genirq: generic chip: Add linear irq domain support

Quoted 11 lines · Grant Likely
Quoted 11 lines
include/linux/irq.h +2 -0
@@ -678,6 +678,8 @@ struct irq_chip_type {
  * @wake_active:	Interrupt is marked as an wakeup from suspend source
  * @num_ct:		Number of available irq_chip_type instances (usually 1)
  * @private:		Private data for non generic chip callbacks
+ * @installed:		bitfield to denote installed interrupts
+ * @domain:		irq domain pointer
  * @list:		List head for keeping track of instances
  * @chip_types:		Array of interrupt irq_chip_types
  *
Quoted 8 lines

include/linux/irq.h struct irq_chip_generic { · line 704

Quoted 5 lines
include/linux/irq.h +1 -0
@@ -699,6 +701,8 @@ struct irq_chip_generic {
 	u32			wake_active;
 	unsigned int		num_ct;
 	void			*private;
+	unsigned long		installed;

This is probably something that the irqdomain should be keeping track of internally, but that's an issue for a separate patch series.

Ok. I just need access to that information, so I can figure out if it's the first irq of the chip which gets mapped. I need this for initializing the mask cache.

Quoted 7 lines

[...]

include/linux/irq.h

Quoted 3 lines
include/linux/irq.h +3 -0
+struct irq_domain_ops irq_generic_chip_ops = {
+	.map = irq_map_generic_chip,
+	.xlate = irq_domain_xlate_onecell,

As discussed on IRC, should use onetwocell here for greater compatibility with existing bindings.

Changed that.

Thanks,

tglx

#67 lore raw

[tip:irq/core] genirq: Generic chip: Add linear irq domain support

In reply to Thomas Gleixner
genirq: Generic chip: Add linear irq domain support
Commit-ID:  088f40b7b027dad6519712ff224a5798dd62a204
Gitweb:     http://git.kernel.org/tip/088f40b7b027dad6519712ff224a5798dd62a204
Author:     Thomas Gleixner <[email protected]>
AuthorDate: Mon, 6 May 2013 14:30:27 +0000
Committer:  Thomas Gleixner <[email protected]>
CommitDate: Wed, 29 May 2013 10:57:11 +0200

genirq: Generic chip: Add linear irq domain support

Provide infrastructure for irq chip implementations which work on linear irq domains.

- Interface to allocate multiple generic chips which are associated to the irq domain.

- Interface to get the generic chip pointer for a particular hardware interrupt in the domain.

- irq domain mapping function to install the chip for a particular interrupt.

Note: This lacks a removal function for now.

[ Sebastian Hesselbarth: Mask cache and pointer math fixups ]

Cc Thomas Petazzoni <[email protected]>
Cc Andrew Lunn <[email protected]>
Cc Russell King - ARM Linux <[email protected]>
Cc Jason Cooper <[email protected]>
Cc Arnd Bergmann <[email protected]>
Cc Jean-Francois Moine <[email protected]>
Cc Rob Herring <[email protected]>
Cc Jason Gunthorpe <[email protected]>
Cc Gregory Clement <[email protected]>
Cc Gerlando Falauto <[email protected]>
Cc Rob Landley <[email protected]>
Cc Maxime Ripard <[email protected]>
Cc Ezequiel Garcia <[email protected]>
Cc Sebastian Hesselbarth <[email protected]>
---
 include/linux/irq.h       |  30 ++++++++
 include/linux/irqdomain.h |  12 +++
 kernel/irq/generic-chip.c | 187 ++++++++++++++++++++++++++++++++++++++++++++--
 kernel/irq/irqdomain.c    |   6 --
 4 files changed, 223 insertions(+), 12 deletions(-)
include/linux/irq.h +30 -0
index ab8169f..af7052c 100644
@@ -678,6 +678,8 @@ struct irq_chip_type {
  * @wake_active:	Interrupt is marked as an wakeup from suspend source
  * @num_ct:		Number of available irq_chip_type instances (usually 1)
  * @private:		Private data for non generic chip callbacks
+ * @installed:		bitfield to denote installed interrupts
+ * @domain:		irq domain pointer
  * @list:		List head for keeping track of instances
  * @chip_types:		Array of interrupt irq_chip_types
  *
@@ -699,6 +701,8 @@ struct irq_chip_generic {
 	u32			wake_active;
 	unsigned int		num_ct;
 	void			*private;
+	unsigned long		installed;
+	struct irq_domain	*domain;
 	struct list_head	list;
 	struct irq_chip_type	chip_types[0];
 };
@@ -719,6 +723,24 @@ enum irq_gc_flags {
 	IRQ_GC_NO_MASK			= 1 << 3,
 };
 
+/*
+ * struct irq_domain_chip_generic - Generic irq chip data structure for irq domains
+ * @irqs_per_chip:	Number of interrupts per chip
+ * @num_chips:		Number of chips
+ * @irq_flags_to_set:	IRQ* flags to set on irq setup
+ * @irq_flags_to_clear:	IRQ* flags to clear on irq setup
+ * @gc_flags:		Generic chip specific setup flags
+ * @gc:			Array of pointers to generic interrupt chips
+ */
+struct irq_domain_chip_generic {
+	unsigned int		irqs_per_chip;
+	unsigned int		num_chips;
+	unsigned int		irq_flags_to_clear;
+	unsigned int		irq_flags_to_set;
+	enum irq_gc_flags	gc_flags;
+	struct irq_chip_generic	*gc[0];
+};
+
 /* Generic chip callback functions */
 void irq_gc_noop(struct irq_data *d);
 void irq_gc_mask_disable_reg(struct irq_data *d);
@@ -742,6 +764,14 @@ int irq_setup_alt_chip(struct irq_data *d, unsigned int type);
 void irq_remove_generic_chip(struct irq_chip_generic *gc, u32 msk,
 			     unsigned int clr, unsigned int set);
 
+struct irq_chip_generic *irq_get_domain_generic_chip(struct irq_domain *d, unsigned int hw_irq);
+int irq_alloc_domain_generic_chips(struct irq_domain *d, int irqs_per_chip,
+				   int num_ct, const char *name,
+				   irq_flow_handler_t handler,
+				   unsigned int clr, unsigned int set,
+				   enum irq_gc_flags flags);
+
+
 static inline struct irq_chip_type *irq_data_get_chip_type(struct irq_data *d)
 {
 	return container_of(d->chip, struct irq_chip_type, chip);
include/linux/irqdomain.h +12 -0
index 0d5b17b..ba2c708 100644
@@ -66,6 +66,10 @@ struct irq_domain_ops {
 		     unsigned long *out_hwirq, unsigned int *out_type);
 };
 
+extern struct irq_domain_ops irq_generic_chip_ops;
+
+struct irq_domain_chip_generic;
+
 /**
  * struct irq_domain - Hardware interrupt number translation object
  * @link: Element in global irq_domain list.
@@ -109,8 +113,16 @@ struct irq_domain {
 
 	/* Optional device node pointer */
 	struct device_node *of_node;
+	/* Optional pointer to generic interrupt chips */
+	struct irq_domain_chip_generic *gc;
 };
 
+#define IRQ_DOMAIN_MAP_LEGACY 0 /* driver allocated fixed range of irqs.
+				 * ie. legacy 8259, gets irqs 1..15 */
+#define IRQ_DOMAIN_MAP_NOMAP 1 /* no fast reverse mapping */
+#define IRQ_DOMAIN_MAP_LINEAR 2 /* linear map of interrupts */
+#define IRQ_DOMAIN_MAP_TREE 3 /* radix tree */
+
 #ifdef CONFIG_IRQ_DOMAIN
 struct irq_domain *irq_domain_add_simple(struct device_node *of_node,
 					 unsigned int size,
kernel/irq/generic-chip.c +181 -6

229 lines not on this page · raw message

kernel/irq/irqdomain.c +0 -6
index 5a83dde..1db9e70 100644
@@ -16,12 +16,6 @@
 #include <linux/smp.h>
 #include <linux/fs.h>
 
-#define IRQ_DOMAIN_MAP_LEGACY 0 /* driver allocated fixed range of irqs.
-				 * ie. legacy 8259, gets irqs 1..15 */
-#define IRQ_DOMAIN_MAP_NOMAP 1 /* no fast reverse mapping */
-#define IRQ_DOMAIN_MAP_LINEAR 2 /* linear map of interrupts */
-#define IRQ_DOMAIN_MAP_TREE 3 /* radix tree */
-
 static LIST_HEAD(irq_domain_list);
 static DEFINE_MUTEX(irq_domain_mutex);
 
#69 lore raw

Re: [patch 8/8] irqchip: sun4i: Convert to generic irq chip

Quoted 71 lines · Thomas Gleixner

Proof of concept patch to demonstrate the new irqdomain support for the generic irq chip. Untested !!

---
 drivers/irqchip/irq-sun4i.c |  100 ++++++++++++--------------------------------
 1 file changed, 29 insertions(+), 71 deletions(-)

Index: linux-2.6/drivers/irqchip/irq-sun4i.c
===================================================================

drivers/irqchip/irq-sun4i.c +3 -44
@@ -32,70 +32,43 @@
 #define SUN4I_IRQ_FIQ_PENDING_REG(x)	(0x20 + 0x4 * x)
 #define SUN4I_IRQ_ENABLE_REG(x)		(0x40 + 0x4 * x)
 #define SUN4I_IRQ_MASK_REG(x)		(0x50 + 0x4 * x)
+#define SUN4I_NUM_CHIPS			3
+#define SUN4I_IRQS_PER_CHIP		32
 
 static void __iomem *sun4i_irq_base;
 static struct irq_domain *sun4i_irq_domain;
 
 static asmlinkage void __exception_irq_entry sun4i_handle_irq(struct pt_regs *regs);
 
-void sun4i_irq_ack(struct irq_data *irqd)
+static int __init sun4i_init_domain_chips(void)
 {
-	unsigned int irq = irqd_to_hwirq(irqd);
-	unsigned int irq_off = irq % 32;
-	int reg = irq / 32;
-	u32 val;
-
-	val = readl(sun4i_irq_base + SUN4I_IRQ_PENDING_REG(reg));
-	writel(val | (1 << irq_off),
-	       sun4i_irq_base + SUN4I_IRQ_PENDING_REG(reg));
-}
-
-static void sun4i_irq_mask(struct irq_data *irqd)
-{
-	unsigned int irq = irqd_to_hwirq(irqd);
-	unsigned int irq_off = irq % 32;
-	int reg = irq / 32;
-	u32 val;
-
-	val = readl(sun4i_irq_base + SUN4I_IRQ_ENABLE_REG(reg));
-	writel(val & ~(1 << irq_off),
-	       sun4i_irq_base + SUN4I_IRQ_ENABLE_REG(reg));
-}
-
-static void sun4i_irq_unmask(struct irq_data *irqd)
-{
-	unsigned int irq = irqd_to_hwirq(irqd);
-	unsigned int irq_off = irq % 32;
-	int reg = irq / 32;
-	u32 val;
-
-	val = readl(sun4i_irq_base + SUN4I_IRQ_ENABLE_REG(reg));
-	writel(val | (1 << irq_off),
-	       sun4i_irq_base + SUN4I_IRQ_ENABLE_REG(reg));
-}
-
-static struct irq_chip sun4i_irq_chip = {
-	.name		= "sun4i_irq",
-	.irq_ack	= sun4i_irq_ack,
-	.irq_mask	= sun4i_irq_mask,
-	.irq_unmask	= sun4i_irq_unmask,
-};
-
-static int sun4i_irq_map(struct irq_domain *d, unsigned int virq,
-			 irq_hw_number_t hw)

drivers/irqchip/irq-sun4i.c line 49

Quoted 10 lines
drivers/irqchip/irq-sun4i.c +5 -5
-{
-	irq_set_chip_and_handler(virq, &sun4i_irq_chip,
-				 handle_level_irq);
-	set_irq_flags(virq, IRQF_VALID | IRQF_PROBE);
-
+	unsigned int clr = IRQ_NOREQUEST | IRQ_NOPROBE | IRQ_NOAUTOEN;
+	struct irq_chip_generic *gc;
+	int i, ret, base = 0;
+
+	ret = irq_alloc_domain_generic_chips(d, SUN4I_IRQS_PER_CHIP, 1,

1 should be SUN4I_NUM_CHIPS.

drivers/irqchip/irq-sun4i.c line 56

Quoted 7 lines
drivers/irqchip/irq-sun4i.c +7 -0
+					     "sun4i_irq", handle_level_irq,
+					     clr, 0, IRQ_GC_INIT_MASK_CACHE);
+	if (ret)
+		return ret;
+
+	for (i = 0; i < SUN4I_NUM_CHIPS; i++, base += SUN4I_IRQS_PER_CHIP) {
+		gc = irq_get_domain_generic_chip(sun4i_irq_domain, base);

Perhaps this could be an iterator macro.

Quoted 46 lines
drivers/irqchip/irq-sun4i.c +12 -20
+		gc->reg_base = sun4i_irq_base;
+		gc->chip_types[0].regs.mask = SUN4I_IRQ_ENABLE_REG(i);
+		gc->chip_types[0].regs.ack = SUN4I_IRQ_PENDING_REG(i);
+		gc->chip_types[0].chip.mask = irq_gc_mask_clr_bit;
+		gc->chip_types[0].chip.ack = irq_gc_ack_set_bit;
+		gc->chip_types[0].chip.unmask = irq_gc_mask_set_bit;
+
+		/* Disable, mask and clear all pending interrupts */
+		writel(0, sun4i_irq_base + SUN4I_IRQ_ENABLE_REG(i));
+		writel(0, sun4i_irq_base + SUN4I_IRQ_MASK_REG(i));
+		writel(0xffffffff, sun4i_irq_base + SUN4I_IRQ_PENDING_REG(i));
+	}
 	return 0;
 }
 
-static struct irq_domain_ops sun4i_irq_ops = {
-	.map = sun4i_irq_map,
-	.xlate = irq_domain_xlate_onecell,
-};
-
 static int __init sun4i_of_init(struct device_node *node,
 				struct device_node *parent)
 {
@@ -104,21 +77,6 @@ static int __init sun4i_of_init(struct d
 		panic("%s: unable to map IC registers\n",
 			node->full_name);
 
-	/* Disable all interrupts */
-	writel(0, sun4i_irq_base + SUN4I_IRQ_ENABLE_REG(0));
-	writel(0, sun4i_irq_base + SUN4I_IRQ_ENABLE_REG(1));
-	writel(0, sun4i_irq_base + SUN4I_IRQ_ENABLE_REG(2));
-
-	/* Mask all the interrupts */
-	writel(0, sun4i_irq_base + SUN4I_IRQ_MASK_REG(0));
-	writel(0, sun4i_irq_base + SUN4I_IRQ_MASK_REG(1));
-	writel(0, sun4i_irq_base + SUN4I_IRQ_MASK_REG(2));
-
-	/* Clear all the pending interrupts */
-	writel(0xffffffff, sun4i_irq_base + SUN4I_IRQ_PENDING_REG(0));
-	writel(0xffffffff, sun4i_irq_base + SUN4I_IRQ_PENDING_REG(1));
-	writel(0xffffffff, sun4i_irq_base + SUN4I_IRQ_PENDING_REG(2));
-
 	/* Enable protection mode */
 	writel(0x01, sun4i_irq_base + SUN4I_IRQ_PROTECTION_REG);
 

drivers/irqchip/irq-sun4i.c static int __init sun4i_of_init(struct d · line 86

Quoted 3 lines
drivers/irqchip/irq-sun4i.c +0 -0
@@ -126,8 +84,8 @@ static int __init sun4i_of_init(struct d
 	writel(0x00, sun4i_irq_base + SUN4I_IRQ_NMI_CTRL_REG);
 
 	sun4i_irq_domain = irq_domain_add_linear(node, 3 * 32,

s/3 * 32/SUN4I_NUM_CHIPS * SUN4I_IRQS_PER_CHIP/

Quoted 13 lines
drivers/irqchip/irq-sun4i.c +2 -2
-						 &sun4i_irq_ops, NULL);
-	if (!sun4i_irq_domain)
+						 &irq_generic_chip_ops, NULL);
+	if (!sun4i_irq_domain || sun4i_init_domain_chips())
 		panic("%s: unable to create IRQ domain\n", node->full_name);
 
 	set_handle_irq(sun4i_handle_irq);

_______________________________________________
devicetree-discuss mailing list
devicetree-discuss at lists.ozlabs.org
https://lists.ozlabs.org/listinfo/devicetree-discuss

#70 lore raw

PATCH 1/1 irq-sun4i: Fix trivial build errors

In reply to Thomas Gleixner
Cc Thomas Gleixner <[email protected]>
Cc Russell King - ARM Linux <[email protected]>
Cc Grant Likely <[email protected]>
Cc Rob Herring <[email protected]>
Cc Rob Landley <[email protected]>
Cc Arnd Bergmann <[email protected]>
Cc Jason Cooper <[email protected]>
Cc Andrew Lunn <[email protected]>
Cc Jason Gunthorpe <[email protected]>
Cc Thomas Petazzoni <[email protected]>
Cc Gregory Clement <[email protected]>
Cc Ezequiel Garcia <[email protected]>
Cc Maxime Ripard <[email protected]>
Cc Jean-Francois Moine <[email protected]>
Cc Gerlando Falauto <[email protected]>
Cc Uwe Kleine-Koenig <[email protected]>
Cc devicetree-discuss at lists.ozlabs.org
Cc linux-doc at vger.kernel.org
Cc linux-arm-kernel at lists.infradead.org
Cc linux-kernel at vger.kernel.org
---
 drivers/irqchip/irq-sun4i.c | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)
drivers/irqchip/irq-sun4i.c +5 -4
index 0191f2c..cab36b1 100644
@@ -46,7 +46,8 @@ static int __init sun4i_init_domain_chips(void)
 	struct irq_chip_generic *gc;
 	int i, ret, base = 0;
 
-	ret = irq_alloc_domain_generic_chips(d, SUN4I_IRQS_PER_CHIP, 1,
+	ret = irq_alloc_domain_generic_chips(sun4i_irq_domain,
+					     SUN4I_IRQS_PER_CHIP, 1,
 					     "sun4i_irq", handle_level_irq,
 					     clr, 0, IRQ_GC_INIT_MASK_CACHE);
 	if (ret)
@@ -57,9 +58,9 @@ static int __init sun4i_init_domain_chips(void)
 		gc->reg_base = sun4i_irq_base;
 		gc->chip_types[0].regs.mask = SUN4I_IRQ_ENABLE_REG(i);
 		gc->chip_types[0].regs.ack = SUN4I_IRQ_PENDING_REG(i);
-		gc->chip_types[0].chip.mask = irq_gc_mask_clr_bit;
-		gc->chip_types[0].chip.ack = irq_gc_ack_set_bit;
-		gc->chip_types[0].chip.unmask = irq_gc_mask_set_bit;
+		gc->chip_types[0].chip.irq_mask = irq_gc_mask_clr_bit;
+		gc->chip_types[0].chip.irq_ack = irq_gc_ack_set_bit;
+		gc->chip_types[0].chip.irq_unmask = irq_gc_mask_set_bit;
 
 		/* Disable, mask and clear all pending interrupts */
 		writel(0, sun4i_irq_base + SUN4I_IRQ_ENABLE_REG(i));
Signature · 1 line
1.8.1.2
Acked-by on v1 8/8
#71 lore raw

Re: [patch 8/8] irqchip: sun4i: Convert to generic irq chip

In reply to Thomas Gleixner

Hi Thomas,

Le 06/05/2013 16:30, Thomas Gleixner a ?crit :

Quoted 5 lines

Proof of concept patch to demonstrate the new irqdomain support for the generic irq chip. Untested !!

I just tested it, and with the minor patch I sent, it's working fine.

You can add my

Thanks,
Maxime

Signature · 3 lines
Maxime Ripard, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com
#72 lore raw

Re: [patch 8/8] irqchip: sun4i: Convert to generic irq chip

Le 12/05/2013 16:08, Maxime Ripard a ?crit :

Quoted 6 lines

Hi Thomas,

Le 06/05/2013 16:30, Thomas Gleixner a ?crit :

Quoted 3 lines

Proof of concept patch to demonstrate the new irqdomain support for the generic irq chip. Untested !!

Quoted 7 lines
Quoted 2 lines

I just tested it, and with the minor patch I sent, it's working fine.

You can add my

After addressing the comments RobH already made that is.

Maxime

Signature · 3 lines
Maxime Ripard, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com
#73 lore raw

Re: [patch 0/8] genirq: Support for irq domains in generic irq chip - V2

In reply to Thomas Gleixner

Hi Thomas,

Quoted 11 lines · Thomas Gleixner

Changes vs. V1:

- Fixed the generic chip pointer thinko (Sebastian Hesselbarth)

- Proper support for mask cache

- Read mask hardware only for the first map of an generic chip instance

- sun4i prefix irq functions proper

[removed all other individual recipients to avoid useless garbage],

thanks for taking care of this. Sorry for the late response, I have been away for a whole week with no email access.

I understand you are providing generic changes and it would be up to us to update individual drivers so to take advantage of those semplifications. Please forgive me if I got this all wrong though.

I've got a few questions then:

1) What happened to your initial proposal of renaming gc->mask_cache to gc->shared_mask_cache? Does that mean you intend to keep compatibility so individual drivers can be updated one at a time?

2) What shall we do with such potential patches for individual drivers, given the dependencies on your work? Wait until yours get merged?

3) [a bit OT] anyone got a clue why I cannot see my previous patch series on patchwork.kernel.org? Neither can I find from within http://marc.info/?l=linux-kernel.
I looked up '[PATCH v3b 0/9] refactoring for mask_cache' and could not find it, only some replies to the thread were shown.
Patches/messages from the same thread are correctly shown on news.gmane.org under gmane.linux.kernel, though threading is broken on my reader; they will not be found through the web interface for gmane.linux.kernel though. I noticed this was a cross post to [email protected] and [email protected] (which I guess is something that should not be done under any circumstances). And all the above works just fine for -stable as opposed to LKML. Could this cross-posting be the root cause for this whole mess?

Thanks!
Gerlando

#74 lore raw

Re: [patch 0/8] genirq: Support for irq domains in generic irq chip - V2

Quoted 3 lines · Gerlando Falauto

1) What happened to your initial proposal of renaming gc->mask_cache to gc->shared_mask_cache? Does that mean you intend to keep compatibility so individual drivers can be updated one at a time?

Yes.

Quoted 2 lines

2) What shall we do with such potential patches for individual drivers, given the dependencies on your work? Wait until yours get merged?

I'll provide a branch for the SoC folks, which they can pull into their tree, so we don't have cross tree dependencies.

Quoted 1 line

Could this cross-posting be the root cause for this whole mess?

No idea, really.

Thanks,

tglx

#75 lore raw

Re: [patch 0/8] genirq: Support for irq domains in generic irq chip - V2

In reply to Thomas Gleixner

Hi Thomas, Sebastian,

I see these changes made it to 3.11.
AFAICT though, 3.10.9 still has the original bug (the one that got me to write the patch for handling separate mask registers) and I am bit confused as to how to integrate that back into 3.10 (or any previous affected kernels, as they deserve a fix as well!).

The way I understand it, any mainstream patch would be based on this work, which is not available on previous kernels. And I guess backporting the whole thing would be overkill.
So I believe the only way to fix it on older kernels would be to write one (or more) minimal version-specific patch series.
But then I wonder: would that be acceptable material for linux-stable?

Please correct me if I'm totally wrong here. I'm willing to help & test but I need directions.

Thanks,
Gerlando

Quoted 16 lines · Thomas Gleixner

Changes vs. V1:

- Fixed the generic chip pointer thinko (Sebastian Hesselbarth)

- Proper support for mask cache

- Read mask hardware only for the first map of an generic chip instance

- sun4i prefix irq functions proper

Thanks,

tglx

author maintainer
#76 lore raw

PATCH v2 2/5 ARM: dove: add DT parsing for legacy mv643xx_eth

In reply to Sebastian Hesselbarth

To allow to move to orion irqchip driver, existing legacy devices have to map their irqs. This patch adds init code to map the corresponding irqs. It will vanish as soon as there is true device tree support for mv643xx_eth.

---
Changelog:
v1->v2:
- split off DT changes (Suggested by Jason Cooper)

Cc Grant Likely <[email protected]>
Cc Rob Herring <[email protected]>
Cc Rob Landley <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc Russell King <[email protected]>
Cc Arnd Bergmann <[email protected]>
Cc Jason Cooper <[email protected]>
Cc Andrew Lunn <[email protected]>
Cc Jason Gunthorpe <[email protected]>
Cc Thomas Petazzoni <[email protected]>
Cc Gregory Clement <[email protected]>
Cc Ezequiel Garcia <[email protected]>
Cc Jean-Francois Moine <[email protected]>
Cc devicetree-discuss at lists.ozlabs.org
Cc linux-doc at vger.kernel.org
Cc linux-arm-kernel at lists.infradead.org
Cc linux-kernel at vger.kernel.org
---
 arch/arm/mach-dove/board-dt.c |   31 ++++++++++++++++++++++++++++++-
 1 file changed, 30 insertions(+), 1 deletion(-)
arch/arm/mach-dove/board-dt.c +30 -1
index fbde1dd..9df6dd7 100644
@@ -12,6 +12,7 @@
 #include <linux/clk-provider.h>
 #include <linux/clk/mvebu.h>
 #include <linux/of.h>
+#include <linux/of_irq.h>
 #include <linux/of_platform.h>
 #include <linux/platform_data/usb-ehci-orion.h>
 #include <asm/hardware/cache-tauros2.h>
@@ -57,6 +58,34 @@ static struct mv643xx_eth_platform_data dove_dt_ge00_data = {
 	.phy_addr = MV643XX_ETH_PHY_ADDR_DEFAULT,
 };
 
+#define DOVE_GE00_PHYS_BASE	0xf1070000
+
+static void __init dove_legacy_ge00_init(void)
+{
+	struct device_node *np = of_find_compatible_node(NULL, NULL,
+					 "marvell,mv643xx-eth-block");
+	int irq_sum, irq_err;
+
+	if (!np)
+		return;
+
+	irq_sum = irq_of_parse_and_map(np, 0);
+	if (!irq_sum) {
+		pr_err("%s: missing sum irq\n", np->full_name);
+		return;
+	}
+
+	irq_err = irq_of_parse_and_map(np, 1);
+	if (!irq_err) {
+		pr_err("%s: missing err irq\n", np->full_name);
+		return;
+	}
+
+	/* legacy ge00_init wants phys base */
+	orion_ge00_init(&dove_dt_ge00_data, DOVE_GE00_PHYS_BASE,
+			irq_sum, irq_err, 1600);
+}
+
 static void __init dove_dt_init(void)
 {
 	pr_info("Dove 88AP510 SoC\n");
@@ -70,7 +99,7 @@ static void __init dove_dt_init(void)
 	dove_of_clk_init();
 
 	/* Internal devices not ported to DT yet */
-	dove_ge00_init(&dove_dt_ge00_data);
+	dove_legacy_ge00_init();
 	dove_pcie_init(1, 1);
 
 	of_platform_populate(NULL, of_default_bus_match_table, NULL, NULL);
Signature · 1 line
1.7.10.4
maintainer
#77 lore raw

Re: [PATCH v2 2/5] ARM: dove: add DT parsing for legacy mv643xx_eth

Quoted 65 lines · Sebastian Hesselbarth

To allow to move to orion irqchip driver, existing legacy devices have to map their irqs. This patch adds init code to map the corresponding irqs. It will vanish as soon as there is true device tree support for mv643xx_eth.

---
Changelog:
v1->v2:
- split off DT changes (Suggested by Jason Cooper)

Cc Grant Likely <[email protected]>
Cc Rob Herring <[email protected]>
Cc Rob Landley <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc Russell King <[email protected]>
Cc Arnd Bergmann <[email protected]>
Cc Jason Cooper <[email protected]>
Cc Andrew Lunn <[email protected]>
Cc Jason Gunthorpe <[email protected]>
Cc Thomas Petazzoni <[email protected]>
Cc Gregory Clement <[email protected]>
Cc Ezequiel Garcia <[email protected]>
Cc Jean-Francois Moine <[email protected]>
Cc devicetree-discuss at lists.ozlabs.org
Cc linux-doc at vger.kernel.org
Cc linux-arm-kernel at lists.infradead.org
Cc linux-kernel at vger.kernel.org
---
 arch/arm/mach-dove/board-dt.c |   31 ++++++++++++++++++++++++++++++-
 1 file changed, 30 insertions(+), 1 deletion(-)
arch/arm/mach-dove/board-dt.c +18 -0
index fbde1dd..9df6dd7 100644
@@ -12,6 +12,7 @@
 #include <linux/clk-provider.h>
 #include <linux/clk/mvebu.h>
 #include <linux/of.h>
+#include <linux/of_irq.h>
 #include <linux/of_platform.h>
 #include <linux/platform_data/usb-ehci-orion.h>
 #include <asm/hardware/cache-tauros2.h>
@@ -57,6 +58,34 @@ static struct mv643xx_eth_platform_data dove_dt_ge00_data = {
 	.phy_addr = MV643XX_ETH_PHY_ADDR_DEFAULT,
 };
 
+#define DOVE_GE00_PHYS_BASE	0xf1070000
+
+static void __init dove_legacy_ge00_init(void)
+{
+	struct device_node *np = of_find_compatible_node(NULL, NULL,
+					 "marvell,mv643xx-eth-block");
+	int irq_sum, irq_err;
+
+	if (!np)
+		return;
+
+	irq_sum = irq_of_parse_and_map(np, 0);
+	if (!irq_sum) {
+		pr_err("%s: missing sum irq\n", np->full_name);
+		return;
+	}
+

arch/arm/mach-dove/board-dt.c static struct mv643xx_eth_platform_data dove_dt_ge00_data = { · line 87

Quoted 10 lines
arch/arm/mach-dove/board-dt.c +10 -0
+	irq_err = irq_of_parse_and_map(np, 1);
+	if (!irq_err) {
+		pr_err("%s: missing err irq\n", np->full_name);
+		return;
+	}
+
+	/* legacy ge00_init wants phys base */
+	orion_ge00_init(&dove_dt_ge00_data, DOVE_GE00_PHYS_BASE,
+			irq_sum, irq_err, 1600);
+}

Hi Sebastian

I know the above code is throw away, but it might help with getting Kirkwood, Orion5x, mv78xx00 supported if we refactor this code and move most of it into plat-orion/common.c. I could imaging a function

orion_ge00_irq_init(struct mv643xx_eth_platform_data *eth_data,
unsigned long mapbase, unsigned int tx_csum_limit)

which does the irq lookup and then calls orion_ge00_init().

Jason: what is the status of the ethernet driver conversion to DT? Will it get merged this week, or is it material for the next merge window?

Andrew

author maintainer
#78 lore raw

Re: [PATCH v2 2/5] ARM: dove: add DT parsing for legacy mv643xx_eth

Quoted 79 lines · Andrew Lunn
Quoted 65 lines · Sebastian Hesselbarth

To allow to move to orion irqchip driver, existing legacy devices have to map their irqs. This patch adds init code to map the corresponding irqs. It will vanish as soon as there is true device tree support for mv643xx_eth.

---
Changelog:
v1->v2:
- split off DT changes (Suggested by Jason Cooper)

Cc Grant Likely<[email protected]>
Cc Rob Herring<[email protected]>
Cc Rob Landley<[email protected]>
Cc Thomas Gleixner<[email protected]>
Cc Russell King<[email protected]>
Cc Arnd Bergmann<[email protected]>
Cc Jason Cooper<[email protected]>
Cc Andrew Lunn<[email protected]>
Cc Jason Gunthorpe<[email protected]>
Cc Thomas Petazzoni<[email protected]>
Cc Gregory Clement<[email protected]>
Cc Ezequiel Garcia<[email protected]>
Cc Jean-Francois Moine<[email protected]>
Cc devicetree-discuss at lists.ozlabs.org
Cc linux-doc at vger.kernel.org
Cc linux-arm-kernel at lists.infradead.org
Cc linux-kernel at vger.kernel.org
---
  arch/arm/mach-dove/board-dt.c |   31 ++++++++++++++++++++++++++++++-
  1 file changed, 30 insertions(+), 1 deletion(-)
arch/arm/mach-dove/board-dt.c +18 -0
index fbde1dd..9df6dd7 100644
@@ -12,6 +12,7 @@
  #include<linux/clk-provider.h>
  #include<linux/clk/mvebu.h>
  #include<linux/of.h>
+#include<linux/of_irq.h>
  #include<linux/of_platform.h>
  #include<linux/platform_data/usb-ehci-orion.h>
  #include<asm/hardware/cache-tauros2.h>
@@ -57,6 +58,34 @@ static struct mv643xx_eth_platform_data dove_dt_ge00_data = {
  	.phy_addr = MV643XX_ETH_PHY_ADDR_DEFAULT,
  };
 
+#define DOVE_GE00_PHYS_BASE	0xf1070000
+
+static void __init dove_legacy_ge00_init(void)
+{
+	struct device_node *np = of_find_compatible_node(NULL, NULL,
+					 "marvell,mv643xx-eth-block");
+	int irq_sum, irq_err;
+
+	if (!np)
+		return;
+
+	irq_sum = irq_of_parse_and_map(np, 0);
+	if (!irq_sum) {
+		pr_err("%s: missing sum irq\n", np->full_name);
+		return;
+	}
+

arch/arm/mach-dove/board-dt.c static struct mv643xx_eth_platform_data dove_dt_ge00_data = { · line 87

Quoted 10 lines
arch/arm/mach-dove/board-dt.c +10 -0
+	irq_err = irq_of_parse_and_map(np, 1);
+	if (!irq_err) {
+		pr_err("%s: missing err irq\n", np->full_name);
+		return;
+	}
+
+	/* legacy ge00_init wants phys base */
+	orion_ge00_init(&dove_dt_ge00_data, DOVE_GE00_PHYS_BASE,
+			irq_sum, irq_err, 1600);
+}

Hi Sebastian

Quoted 8 lines

I know the above code is throw away, but it might help with getting Kirkwood, Orion5x, mv78xx00 supported if we refactor this code and move most of it into plat-orion/common.c. I could imaging a function

orion_ge00_irq_init(struct mv643xx_eth_platform_data *eth_data,
unsigned long mapbase, unsigned int tx_csum_limit)

which does the irq lookup and then calls orion_ge00_init().

Andrew,

sure. Feel free to move legacy code to where all SoCs can use it.

Quoted 3 lines

Jason: what is the status of the ethernet driver conversion to DT? Will it get merged this week, or is it material for the next merge window?

What I remember from Florian's work it was close to be released. I never saw Florian sending an update to his patches, maybe he ran out of time?

I guess DT timer will be a quick patch when we have all SoCs running on DT irqchip and maybe Florian can update his patches even before we sort out Thomas request. Then we could also merge DT timer within this patch set and get rid of legacy helpers before we even introduce them.

Sebastian

#79 lore raw

Re: [PATCH v2 2/5] ARM: dove: add DT parsing for legacy mv643xx_eth

In reply to Andrew Lunn
Quoted 3 lines · Andrew Lunn

Jason: what is the status of the ethernet driver conversion to DT? Will it get merged this week, or is it material for the next merge window?

You'd have to ask Florian/David Miller. I'm not sure wether David was able to pull that in for v3.10 or not. He was pretty responsive to all of the other changes Florian posted.

thx,

Jason.

A message that is not in the archive (sent off-list or not yet received): 20130504182935.GO31290-u4khhh1J0LxI1Ri9qeTfzeTW4wlIGRCZ@public.gmane.org

#81 lore raw

Re: [PATCH v2 2/5] ARM: dove: add DT parsing for legacy mv643xx_eth

Quoted 9 lines · Jason Cooper
Quoted 3 lines · Andrew Lunn

Jason: what is the status of the ethernet driver conversion to DT? Will it get merged this week, or is it material for the next merge window?

You'd have to ask Florian/David Miller. I'm not sure wether David was able to pull that in for v3.10 or not. He was pretty responsive to all of the other changes Florian posted.

Sorry for the delay, I do not think I will be able to work on these patches in the next few weeks. I would appreciate if Sebastian or yourself could follow up with the latest submission? Sebastian already sent me an updated version which looks good enough to be sent for review

_______________________________________________
devicetree-discuss mailing list devicetree-discuss-uLR06cmDAlY/[email protected]
https://lists.ozlabs.org/listinfo/devicetree-discuss

author maintainer
#82 lore raw

PATCH v2 3/5 ARM: dove: add DT parsing for legacy timer

In reply to Sebastian Hesselbarth

To allow to move to orion irqchip driver, existing legacy devices have to map their irqs. This patch adds init code to map the corresponding irqs. It will vanish as soon as there is true device tree support for orion timer.

---
Changelog:
v1->v2:
- split off DT changes (Suggested by Jason Cooper)

Cc Grant Likely <[email protected]>
Cc Rob Herring <[email protected]>
Cc Rob Landley <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc Russell King <[email protected]>
Cc Arnd Bergmann <[email protected]>
Cc Jason Cooper <[email protected]>
Cc Andrew Lunn <[email protected]>
Cc Jason Gunthorpe <[email protected]>
Cc Thomas Petazzoni <[email protected]>
Cc Gregory Clement <[email protected]>
Cc Ezequiel Garcia <[email protected]>
Cc Jean-Francois Moine <[email protected]>
Cc devicetree-discuss at lists.ozlabs.org
Cc linux-doc at vger.kernel.org
Cc linux-arm-kernel at lists.infradead.org
Cc linux-kernel at vger.kernel.org
---
 arch/arm/mach-dove/board-dt.c |   42 +++++++++++++++++++++++++++++++++++------
 1 file changed, 36 insertions(+), 6 deletions(-)
arch/arm/mach-dove/board-dt.c +36 -6
index 9df6dd7..cea65b7 100644
@@ -12,6 +12,7 @@
 #include <linux/clk-provider.h>
 #include <linux/clk/mvebu.h>
 #include <linux/of.h>
+#include <linux/of_address.h>
 #include <linux/of_irq.h>
 #include <linux/of_platform.h>
 #include <linux/platform_data/usb-ehci-orion.h>
@@ -20,6 +21,7 @@
 #include <mach/pm.h>
 #include <plat/common.h>
 #include <plat/irq.h>
+#include <plat/time.h>
 #include "common.h"
 
 /*
@@ -48,10 +50,42 @@ static void __init dove_legacy_clk_init(void)
 			 of_clk_get_from_provider(&clkspec));
 }
 
-static void __init dove_of_clk_init(void)
+#define BRIDGE_INT_TIMER1_CLR	(~0x0004)
+
+static void __init dove_legacy_timer_init(void)
 {
+	struct device_node *np = of_find_compatible_node(NULL, NULL,
+					 "marvell,orion-timer");
+	struct clk *clk;
+	void __iomem *base;
+	unsigned int tclk, irq;
+
+	/* Setup root of clk tree */
 	mvebu_clocks_init();
 	dove_legacy_clk_init();
+
+	if (!np)
+		panic("missing timer node\n");
+
+	base = of_iomap(np, 0);
+	if (!base)
+		panic("%s: missing reg base for timer\n", np->full_name);
+
+	irq = irq_of_parse_and_map(np, 0);
+	if (!irq)
+		panic("%s: missing irq for timer\n", np->full_name);
+
+	clk = of_clk_get(np, 0);
+	if (IS_ERR(clk))
+		panic("%s: missing clock for timer\n", np->full_name);
+
+	clk_prepare_enable(clk);
+	tclk = clk_get_rate(clk);
+	clk_put(clk);
+
+	orion_time_set_base(base);
+	/* legacy timer init gets bridge reg base */
+	orion_time_init(base - 0x300, BRIDGE_INT_TIMER1_CLR, irq, tclk);
 }
 
 static struct mv643xx_eth_platform_data dove_dt_ge00_data = {
@@ -95,9 +129,6 @@ static void __init dove_dt_init(void)
 #endif
 	dove_setup_cpu_mbus();
 
-	/* Setup root of clk tree */
-	dove_of_clk_init();
-
 	/* Internal devices not ported to DT yet */
 	dove_legacy_ge00_init();
 	dove_pcie_init(1, 1);
@@ -112,9 +143,8 @@ static const char * const dove_dt_board_compat[] = {
 
 DT_MACHINE_START(DOVE_DT, "Marvell Dove (Flattened Device Tree)")
 	.map_io		= dove_map_io,
-	.init_early	= dove_init_early,
 	.init_irq	= orion_dt_init_irq,
-	.init_time	= dove_timer_init,
+	.init_time	= dove_legacy_timer_init,
 	.init_machine	= dove_dt_init,
 	.restart	= dove_restart,
 	.dt_compat	= dove_dt_board_compat,
Signature · 1 line
1.7.10.4
author maintainer
#83 lore raw

PATCH v2 4/5 ARM: dove: move DT boards to orion irqchip driver

In reply to Sebastian Hesselbarth

With legacy devices mapping their irqs, we can now switch DT enabled boards to orion irqchip driver.

---
Changelog:
v1->v2:
- split off DT changes (Suggested by Jason Cooper)
- use irqchip_init (Suggested by Jason Gunthorpe)

Cc Grant Likely <[email protected]>
Cc Rob Herring <[email protected]>
Cc Rob Landley <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc Russell King <[email protected]>
Cc Arnd Bergmann <[email protected]>
Cc Jason Cooper <[email protected]>
Cc Andrew Lunn <[email protected]>
Cc Jason Gunthorpe <[email protected]>
Cc Thomas Petazzoni <[email protected]>
Cc Gregory Clement <[email protected]>
Cc Ezequiel Garcia <[email protected]>
Cc Jean-Francois Moine <[email protected]>
Cc devicetree-discuss at lists.ozlabs.org
Cc linux-doc at vger.kernel.org
Cc linux-arm-kernel at lists.infradead.org
Cc linux-kernel at vger.kernel.org
---
 arch/arm/mach-dove/Kconfig    |    1 +
 arch/arm/mach-dove/Makefile   |    4 ++--
 arch/arm/mach-dove/board-dt.c |    4 ++--
 3 files changed, 5 insertions(+), 4 deletions(-)
arch/arm/mach-dove/Kconfig +1 -0
index 36469d8..60d72b4 100644
@@ -22,6 +22,7 @@ config MACH_CM_A510
 
 config MACH_DOVE_DT
 	bool "Marvell Dove Flattened Device Tree"
+	select IRQCHIP_ORION
 	select MVEBU_CLK_CORE
 	select MVEBU_CLK_GATING
 	select REGULATOR
arch/arm/mach-dove/Makefile +2 -2
index 3f0a858..4780f78 100644
@@ -1,5 +1,5 @@
-obj-y				+= common.o addr-map.o irq.o
-obj-$(CONFIG_DOVE_LEGACY)	+= mpp.o
+obj-y				+= common.o addr-map.o
+obj-$(CONFIG_DOVE_LEGACY)	+= irq.o mpp.o
 obj-$(CONFIG_PCI)		+= pcie.o
 obj-$(CONFIG_MACH_DOVE_DB)	+= dove-db-setup.o
 obj-$(CONFIG_MACH_DOVE_DT)	+= board-dt.o
arch/arm/mach-dove/board-dt.c +2 -2
index cea65b7..404c993 100644
@@ -11,6 +11,7 @@
 #include <linux/init.h>
 #include <linux/clk-provider.h>
 #include <linux/clk/mvebu.h>
+#include <linux/irqchip.h>
 #include <linux/of.h>
 #include <linux/of_address.h>
 #include <linux/of_irq.h>
@@ -20,7 +21,6 @@
 #include <asm/mach/arch.h>
 #include <mach/pm.h>
 #include <plat/common.h>
-#include <plat/irq.h>
 #include <plat/time.h>
 #include "common.h"
 
@@ -143,7 +143,7 @@ static const char * const dove_dt_board_compat[] = {
 
 DT_MACHINE_START(DOVE_DT, "Marvell Dove (Flattened Device Tree)")
 	.map_io		= dove_map_io,
-	.init_irq	= orion_dt_init_irq,
+	.init_irq	= irqchip_init,
 	.init_time	= dove_legacy_timer_init,
 	.init_machine	= dove_dt_init,
 	.restart	= dove_restart,
Signature · 1 line
1.7.10.4
author maintainer
#84 lore raw

PATCH v2 5/5 ARM: dove: add DT nodes for irqchip conversion

In reply to Sebastian Hesselbarth

This patch adds required device tree nodes for irqchip conversion on Dove. The nodes are named and layed out to allow seamless migration to true DT enabled drivers as soon as they are available.

---
Changelog:
v1->v2:
- merge DT changes into this patch (Suggested by Jason Cooper)

Cc Grant Likely <[email protected]>
Cc Rob Herring <[email protected]>
Cc Rob Landley <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc Russell King <[email protected]>
Cc Arnd Bergmann <[email protected]>
Cc Jason Cooper <[email protected]>
Cc Andrew Lunn <[email protected]>
Cc Jason Gunthorpe <[email protected]>
Cc Thomas Petazzoni <[email protected]>
Cc Gregory Clement <[email protected]>
Cc Ezequiel Garcia <[email protected]>
Cc Jean-Francois Moine <[email protected]>
Cc devicetree-discuss at lists.ozlabs.org
Cc linux-doc at vger.kernel.org
Cc linux-arm-kernel at lists.infradead.org
Cc linux-kernel at vger.kernel.org
---
 arch/arm/boot/dts/dove.dtsi |   16 +++++++++++++++-
 1 file changed, 15 insertions(+), 1 deletion(-)
arch/arm/boot/dts/dove.dtsi +15 -1
index 6cab468..afb719a 100644
@@ -30,11 +30,18 @@
 			marvell,tauros2-cache-features = <0>;
 		};
 
+		timer: timer {
+			compatible = "marvell,orion-timer";
+			reg = <0x20300 0x20>;
+			interrupts = <0>;
+			clocks = <&core_clk 0>;
+		};
+
 		intc: interrupt-controller {
 			compatible = "marvell,orion-intc";
 			interrupt-controller;
 			#interrupt-cells = <1>;
-			reg = <0x20204 0x04>, <0x20214 0x04>;
+			reg = <0x20200 0x10>, <0x20210 0x10>;
 		};
 
 		core_clk: core-clocks at d0214 {
@@ -202,6 +209,13 @@
 			status = "disabled";
 		};
 
+		gbe: ethernet-controller {
+			compatible = "marvell,mv643xx-eth-block";
+			reg = <0x72000 0x4000>;
+			interrupts = <29>, <30>;
+			status = "disabled";
+		};
+
 		rtc at d8500 {
 			compatible = "marvell,orion-rtc";
 			reg = <0xd8500 0x20>;
Signature · 1 line
1.7.10.4