Skip to contentSkip to conversation

Add OF support to CMT driver - bindings discussion

PATCH New Bastian Hecht 4 participants 13 messages Mar 1, 2013 – Mar 6, 2013

Subsystems (6)
lore
Share & tools
lore
#0 lore raw

cover Add OF support to CMT driver - bindings discussion

Hello,

here in the CMT driver we currently can derive all neccessary hardware info from the device ID and channel ID (there are multiple CMT instances built into one SoC with multiple channels each). This seems like a very intuituve description to me.

If multiple register layouts appear we can add an (optional) regtype property and some parts of the driver need to be rewritten anyway. But the descriptiveness of device ID and channel ID should stay I think.

Will this work?

In the patchset I have added the dts specification to sh73a0-kzm9g-reference.dts to test it. Does this belong to sh73a0.dtsi instead? I wonder because the clock is in the setup code, not board code.

Thanks,

Bastian

Bastian Hecht (3):
clocksource: sh_cmt: Add Device Tree probing
ARM: mach-shmobile: sh73a0: Add CMT DT name to clock list
ARM: mach-shmobile: sh73a0: Setup the timer device CMT10 using DT

 arch/arm/boot/dts/sh73a0-kzm9g-reference.dts |   11 ++++
 arch/arm/mach-shmobile/clock-sh73a0.c        |    1 +
 arch/arm/mach-shmobile/setup-sh73a0.c        |    1 -
 drivers/clocksource/sh_cmt.c                 |   87 ++++++++++++++++++++++----
 4 files changed, 86 insertions(+), 14 deletions(-)
Signature · 1 line
1.7.9.5
#1 lore raw

PATCH 1/3 clocksource: sh_cmt: Add Device Tree probing

We add the capabilty to probe SH CMT timer devices using Device Tree setup.

We can deduce former platform data by the device IDs and channel IDs of our timer instances, so we choose this more intuitive info as our DT properties.

---
 .../bindings/timer/renesas,sh-cmt-timer.txt        |   21 +++++
 drivers/clocksource/sh_cmt.c                       |   94 +++++++++++++++++---
 2 files changed, 102 insertions(+), 13 deletions(-)
 create mode 100644 Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +21 -0
new file mode 100644
index 0000000..e5ad808
@@ -0,0 +1,21 @@
+* Renesas SH Mobile Compare Match Timer
+
+Required properties:
+- compatible : Should be "renesas,cmt"
+- reg : Address and length of the register set for the device
+- interrupts : Timer IRQ
+- renesas,timer-device-id : The ID of the timer device
+- renesas,timer-channel-id : The channel ID of the timer device
+
+Example for CMT10 of the R8A7740 SoC:
+
+	cmt@0xe6138010 {
+		compatible = "renesas,cmt";
+		interrupt-parent = <&intca>;
+		reg = <0xe6138010 0xc>;
+		interrupts = <0x0b00>;
+		renesas,timer-device-id = <1>;
+		renesas,timer-channel-id = <0>;
+		renesas,clocksource-rating = <150>;
+		renesas,clockevent-rating = <150>;
+	};
drivers/clocksource/sh_cmt.c +81 -13

162 lines not on this page · raw message

Signature · 1 line
1.7.9.5
#2 lore raw

PATCH 1/3 clocksource: sh_cmt: Add Device Tree probing

Quoted 7 lines · Bastian Hecht

We add the capabilty to probe SH CMT timer devices using Device Tree setup.

We can deduce former platform data by the device IDs and channel IDs of our timer instances, so we choose this more intuitive info as our DT properties.

I wonder if it makes more sense to attempt to create a more generalized binding, as what you have here is pretty much uniform for all of CMT/TMU/MTU2. Perhaps having a general sh-timer binding would be cleaner, while letting each timer plug the compatible string it cares about.

#3 lore raw

PATCH 1/3 clocksource: sh_cmt: Add Device Tree probing

Hi Paul,

2013/3/4 Paul Mundt <[email protected]>:

Quoted 12 lines · Bastian Hecht
Quoted 7 lines

We add the capabilty to probe SH CMT timer devices using Device Tree setup.

We can deduce former platform data by the device IDs and channel IDs of our timer instances, so we choose this more intuitive info as our DT properties.

I wonder if it makes more sense to attempt to create a more generalized binding, as what you have here is pretty much uniform for all of CMT/TMU/MTU2. Perhaps having a general sh-timer binding would be cleaner, while letting each timer plug the compatible string it cares about.

Yes that sounds good. I'll have a look at the various timers and see if I can come up with something good.

Thanks,

Bastian

#4 lore raw

PATCH 1/3 clocksource: sh_cmt: Add Device Tree probing

In reply to Bastian Hecht

Hello,

I have a couple of comments on the binding and the way it's parsed.

Quoted 20 lines · Bastian Hecht

We add the capabilty to probe SH CMT timer devices using Device Tree setup.

We can deduce former platform data by the device IDs and channel IDs of our timer instances, so we choose this more intuitive info as our DT properties.

---
 .../bindings/timer/renesas,sh-cmt-timer.txt        |   21 +++++
 drivers/clocksource/sh_cmt.c                       |   94 +++++++++++++++++---
 2 files changed, 102 insertions(+), 13 deletions(-)
 create mode 100644 Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +0 -0
new file mode 100644
index 0000000..e5ad808

Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt line 8

Quoted 8 lines
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +8 -0
@@ -0,0 +1,21 @@
+* Renesas SH Mobile Compare Match Timer
+
+Required properties:
+- compatible : Should be "renesas,cmt"
+- reg : Address and length of the register set for the device
+- interrupts : Timer IRQ
+- renesas,timer-device-id : The ID of the timer device
+- renesas,timer-channel-id : The channel ID of the timer device

I'm not familiar with this hardware. Could you give me a basic idea of how it's structured?

Does the device have a single timer that this describes, or does the device have multiple timers, and this selects which one to use?

Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt line 20

Quoted 12 lines
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +12 -0
+
+Example for CMT10 of the R8A7740 SoC:
+
+	cmt at 0xe6138010 {
+		compatible = "renesas,cmt";
+		interrupt-parent = <&intca>;
+		reg = <0xe6138010 0xc>;
+		interrupts = <0x0b00>;
+		renesas,timer-device-id = <1>;
+		renesas,timer-channel-id = <0>;
+		renesas,clocksource-rating = <150>;
+		renesas,clockevent-rating = <150>;

I'm not sure these last two are describing the hardware as such, they look like Linux implementation details. Do these really need to be in the binding?

[...]

Quoted 9 lines
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +9 -0
+static struct sh_timer_config *sh_cmt_parse_dt(struct device *dev)
+{
+	struct sh_timer_config *cfg;
+	struct device_node *np = dev->of_node;
+	const __be32 *prop;
+	int timer_id, channel_id;
+
+	cfg = devm_kzalloc(dev, sizeof(struct sh_timer_config), GFP_KERNEL);
+	if (!cfg) {

Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt

Quoted 10 lines
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +10 -0
+		dev_err(dev, "failed to allocate DT config data\n");
+		return NULL;
+	}
+
+	prop = of_get_property(np, "renesas,timer-device-id", NULL);
+	if (!prop) {
+		dev_err(dev, "device id missing\n");
+		return NULL;
+	}
+	timer_id = be32_to_cpup(prop);

You can use of_property_read_u32 here, e.g.

if (!of_property_read_u32(np, "renesas,timer-device-id", &timer_id) {
	dev_err(dev, "device id missing\n");
	return NULL;
}

Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt

Quoted 7 lines
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +7 -0
+
+	prop = of_get_property(np, "renesas,timer-channel-id", NULL);
+	if (!prop) {
+		dev_err(dev, "channel id missing\n");
+		return NULL;
+	}
+	channel_id = be32_to_cpup(prop);

You can use of_property_read_u32 here too.

I assume there's a sensible range of channel_id values that could be checked before it's used in a calculation below.

Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt

Quoted 13 lines
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +13 -0
+
+	if (timer_id >= ARRAY_SIZE(sh_timer_offset_multiplier)) {
+		dev_err(dev, "unsupported timer id\n");
+		return NULL;
+	}
+
+	cfg->channel_offset = sh_timer_offset_multiplier[timer_id] *
+							(channel_id + 1);
+	cfg->timer_bit = channel_id;
+
+	prop = of_get_property(np, "renesas,clocksource-rating", NULL);
+	if (prop)
+		cfg->clocksource_rating = be32_to_cpup(prop);

You can use of_property_read_u32 here too.

Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt

Quoted 4 lines
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +4 -0
+
+	prop = of_get_property(np, "renesas,clockevent-rating", NULL);
+	if (prop)
+		cfg->clockevent_rating = be32_to_cpup(prop);

And here.

[...]

Thanks,
Mark.

#5 lore raw

PATCH 1/3 clocksource: sh_cmt: Add Device Tree probing

Hi Mark,

2013/3/4 Mark Rutland <[email protected]>:

Quoted 3 lines

Hello,

I have a couple of comments on the binding and the way it's parsed.

Thank you - appreciated!

Quoted 33 lines · Bastian Hecht
Quoted 20 lines

We add the capabilty to probe SH CMT timer devices using Device Tree setup.

We can deduce former platform data by the device IDs and channel IDs of our timer instances, so we choose this more intuitive info as our DT properties.

---
 .../bindings/timer/renesas,sh-cmt-timer.txt        |   21 +++++
 drivers/clocksource/sh_cmt.c                       |   94 +++++++++++++++++---
 2 files changed, 102 insertions(+), 13 deletions(-)
 create mode 100644 Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +0 -0
new file mode 100644
index 0000000..e5ad808

Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt line 8

Quoted 8 lines
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +8 -0
@@ -0,0 +1,21 @@
+* Renesas SH Mobile Compare Match Timer
+
+Required properties:
+- compatible : Should be "renesas,cmt"
+- reg : Address and length of the register set for the device
+- interrupts : Timer IRQ
+- renesas,timer-device-id : The ID of the timer device
+- renesas,timer-channel-id : The channel ID of the timer device

I'm not familiar with this hardware. Could you give me a basic idea of how it's structured?

Quoted 2 lines

Does the device have a single timer that this describes, or does the device have multiple timers, and this selects which one to use?

The CMT timer is built into various SoCs from Renesas, often multiple instances of it. Each timer instance has 6 channels. You can select if they are driven by the main clock or the real time clock (the SoCs include an SH core for data processing) and can use different subscale modes. It features free running mode and can fire events. The clock has up to 48 bits to count. It has a DMA channel and wires enabling the device to be a wakeup source. Hmm don't know what to add more. It's a timer - you can check for overflows, read the counter and so on.

Quoted 7 lines
Quoted 7 lines
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +7 -0
+
+Example for CMT10 of the R8A7740 SoC:
+
+     cmt at 0xe6138010 {
+             compatible = "renesas,cmt";
+             interrupt-parent = <&intca>;
+             reg = <0xe6138010 0xc>;
Quoted 8 lines

Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt line 20

Quoted 5 lines
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +5 -0
+             interrupts = <0x0b00>;
+             renesas,timer-device-id = <1>;
+             renesas,timer-channel-id = <0>;
+             renesas,clocksource-rating = <150>;
+             renesas,clockevent-rating = <150>;

I'm not sure these last two are describing the hardware as such, they look like Linux implementation details. Do these really need to be in the binding?

That's true - they don't describe the hardware. It's a policy how the device should be used. Here: Register it as clocksource *and* eventsource. I don't know if there is a use case when we want only 1 of them.

Quoted 15 lines

[...]

Quoted 13 lines
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +13 -0
+static struct sh_timer_config *sh_cmt_parse_dt(struct device *dev)
+{
+     struct sh_timer_config *cfg;
+     struct device_node *np = dev->of_node;
+     const __be32 *prop;
+     int timer_id, channel_id;
+
+     cfg = devm_kzalloc(dev, sizeof(struct sh_timer_config), GFP_KERNEL);
+     if (!cfg) {
+             dev_err(dev, "failed to allocate DT config data\n");
+             return NULL;
+     }
+
Quoted 8 lines

Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt

Quoted 6 lines
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +6 -0
+     prop = of_get_property(np, "renesas,timer-device-id", NULL);
+     if (!prop) {
+             dev_err(dev, "device id missing\n");
+             return NULL;
+     }
+     timer_id = be32_to_cpup(prop);

You can use of_property_read_u32 here, e.g.

Oh yes nice! I will convert these calls.

Quoted 13 lines
if (!of_property_read_u32(np, "renesas,timer-device-id", &timer_id) {
        dev_err(dev, "device id missing\n");
        return NULL;
}

Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt

Quoted 7 lines
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +7 -0
+
+     prop = of_get_property(np, "renesas,timer-channel-id", NULL);
+     if (!prop) {
+             dev_err(dev, "channel id missing\n");
+             return NULL;
+     }
+     channel_id = be32_to_cpup(prop);
Quoted 4 lines

You can use of_property_read_u32 here too.

I assume there's a sensible range of channel_id values that could be checked before it's used in a calculation below.

Oh true. Jeeez - this very much looks like exploitable code. If someone manages to change the boot config...
Whatever - I'll add a check.

Quoted 7 lines
Quoted 7 lines
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +7 -0
+
+     if (timer_id >= ARRAY_SIZE(sh_timer_offset_multiplier)) {
+             dev_err(dev, "unsupported timer id\n");
+             return NULL;
+     }
+
+     cfg->channel_offset = sh_timer_offset_multiplier[timer_id] *
Quoted 8 lines

Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt

Quoted 6 lines
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +6 -0
+                                                     (channel_id + 1);
+     cfg->timer_bit = channel_id;
+
+     prop = of_get_property(np, "renesas,clocksource-rating", NULL);
+     if (prop)
+             cfg->clocksource_rating = be32_to_cpup(prop);

You can use of_property_read_u32 here too.

Will do.

Quoted 6 lines

Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt

Quoted 4 lines
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +4 -0
+
+     prop = of_get_property(np, "renesas,clockevent-rating", NULL);
+     if (prop)
+             cfg->clockevent_rating = be32_to_cpup(prop);

And here.

Dito.

Quoted 4 lines

[...]

Thanks,
Mark.

Thanks for this productive review!

Bastian

#6 lore raw

PATCH 1/3 clocksource: sh_cmt: Add Device Tree probing

Hi again,

Quoted 21 lines
Quoted 18 lines
Quoted 6 lines
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +0 -0
new file mode 100644
index 0000000..e5ad808

Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt line 8

Quoted 8 lines
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +8 -0
@@ -0,0 +1,21 @@
+* Renesas SH Mobile Compare Match Timer
+
+Required properties:
+- compatible : Should be "renesas,cmt"
+- reg : Address and length of the register set for the device
+- interrupts : Timer IRQ
+- renesas,timer-device-id : The ID of the timer device
+- renesas,timer-channel-id : The channel ID of the timer device

I'm not familiar with this hardware. Could you give me a basic idea of how it's structured?

Quoted 2 lines

Does the device have a single timer that this describes, or does the device have multiple timers, and this selects which one to use?

Quoted 9 lines

The CMT timer is built into various SoCs from Renesas, often multiple instances of it. Each timer instance has 6 channels. You can select if they are driven by the main clock or the real time clock (the SoCs include an SH core for data processing) and can use different subscale modes. It features free running mode and can fire events. The clock has up to 48 bits to count. It has a DMA channel and wires enabling the device to be a wakeup source. Hmm don't know what to add more. It's a timer - you can check for overflows, read the counter and so on.

Unless I've misunderstood, couldn't Linux choose one timer to use completely arbitrarily?

Why do we need to describe which one to use?

Quoted 17 lines
Quoted 16 lines

Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt line 20

Quoted 13 lines
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +12 -0
+
+Example for CMT10 of the R8A7740 SoC:
+
+     cmt at 0xe6138010 {
+             compatible = "renesas,cmt";
+             interrupt-parent = <&intca>;
+             reg = <0xe6138010 0xc>;
+             interrupts = <0x0b00>;
+             renesas,timer-device-id = <1>;
+             renesas,timer-channel-id = <0>;
+             renesas,clocksource-rating = <150>;
+             renesas,clockevent-rating = <150>;

I'm not sure these last two are describing the hardware as such, they look like Linux implementation details. Do these really need to be in the binding?

Quoted 4 lines

That's true - they don't describe the hardware. It's a policy how the device should be used. Here: Register it as clocksource *and* eventsource. I don't know if there is a use case when we want only 1 of them.

I don't think it's a good idea to expose something like this through the binding, as it's strongly tied to the current Linux implementation. If we can always use the device as both, we may as well (other drivers already do). If the quality of the clocksource / timer is variable enough that it's worth describing, maybe there's a better way of describing this variability specifically?

Quoted 25 lines
Quoted 16 lines

[...]

Quoted 13 lines
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +13 -0
+static struct sh_timer_config *sh_cmt_parse_dt(struct device *dev)
+{
+     struct sh_timer_config *cfg;
+     struct device_node *np = dev->of_node;
+     const __be32 *prop;
+     int timer_id, channel_id;
+
+     cfg = devm_kzalloc(dev, sizeof(struct sh_timer_config), GFP_KERNEL);
+     if (!cfg) {
+             dev_err(dev, "failed to allocate DT config data\n");
+             return NULL;
+     }
+
Quoted 8 lines

Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt

Quoted 6 lines
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +6 -0
+     prop = of_get_property(np, "renesas,timer-device-id", NULL);
+     if (!prop) {
+             dev_err(dev, "device id missing\n");
+             return NULL;
+     }
+     timer_id = be32_to_cpup(prop);

You can use of_property_read_u32 here, e.g.

Quoted 6 lines

Oh yes nice! I will convert these calls.

Quoted 4 lines
if (!of_property_read_u32(np, "renesas,timer-device-id", &timer_id) {
        dev_err(dev, "device id missing\n");
        return NULL;
}

I've just realised some of the variables being assigned to aren't u32s. While it'd still be nicer to use of_property_read_u32, you may still need a temporary variable.

Quoted 9 lines
Quoted 9 lines
Quoted 8 lines
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +7 -0
+
+     prop = of_get_property(np, "renesas,timer-channel-id", NULL);
+     if (!prop) {
+             dev_err(dev, "channel id missing\n");
+             return NULL;
+     }
+     channel_id = be32_to_cpup(prop);
Quoted 8 lines
Quoted 4 lines

You can use of_property_read_u32 here too.

I assume there's a sensible range of channel_id values that could be checked before it's used in a calculation below.

Oh true. Jeeez - this very much looks like exploitable code. If someone manages to change the boot config...
Whatever - I'll add a check.

What I meant was it'd be nice to warn the user that the dt doesn't look right so it's easier to debug when something blows up unexpectedly.

[...]

Thanks,
Mark.

#7 lore raw

PATCH 1/3 clocksource: sh_cmt: Add Device Tree probing

Quoted 33 lines · Mark Rutland

Hi again,

Quoted 21 lines
Quoted 18 lines
Quoted 6 lines
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +0 -0
new file mode 100644
index 0000000..e5ad808

Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt line 8

Quoted 8 lines
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +8 -0
@@ -0,0 +1,21 @@
+* Renesas SH Mobile Compare Match Timer
+
+Required properties:
+- compatible : Should be "renesas,cmt"
+- reg : Address and length of the register set for the device
+- interrupts : Timer IRQ
+- renesas,timer-device-id : The ID of the timer device
+- renesas,timer-channel-id : The channel ID of the timer device

I'm not familiar with this hardware. Could you give me a basic idea of how it's structured?

Quoted 2 lines

Does the device have a single timer that this describes, or does the device have multiple timers, and this selects which one to use?

Quoted 9 lines

The CMT timer is built into various SoCs from Renesas, often multiple instances of it. Each timer instance has 6 channels. You can select if they are driven by the main clock or the real time clock (the SoCs include an SH core for data processing) and can use different subscale modes. It features free running mode and can fire events. The clock has up to 48 bits to count. It has a DMA channel and wires enabling the device to be a wakeup source. Hmm don't know what to add more. It's a timer - you can check for overflows, read the counter and so on.

Quoted 5 lines

Unless I've misunderstood, couldn't Linux choose one timer to use completely arbitrarily?

Why do we need to describe which one to use?

Many of these CPUs have multiple timer blocks (CMT, TMU, MTU2) that sit in different power domains, so there is often a desire to prioritize one over another, as it will determine which sleep states are available to us. Beyond that, many of these channels have special behavioural constraints (only certain channels being useable by the PWM block for example, as well as mux issues (certain channels will be tied up with functions we care less about than others).

We have traditionally registered all of the available timer channels as platform devices, and then left it to the clocksource/event rating to determine which to use for what. This was based on the assumption that later on we would do some work on the clockevents layer to dynamically allocate the other channels that weren't being used for anything, but this never happened, and most users managed to get away with simply adopting hrtimers instead.

Perhaps the rating is not a detail we want to expose, but we are still going to need some facility for defining whether a given timer is suitable for a clocksource/clockevent or not. We also can't really encode the rating information in the driver itself, as the placement of the block is wholly SoC dependent.

#8 lore raw

PATCH 1/3 clocksource: sh_cmt: Add Device Tree probing

Hi Paul, hi Mark,

2013/3/6 Paul Mundt <[email protected]>:

Quoted 55 lines · Mark Rutland
Quoted 33 lines

Hi again,

Quoted 21 lines
Quoted 18 lines
Quoted 6 lines
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +0 -0
new file mode 100644
index 0000000..e5ad808

Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt line 8

Quoted 8 lines
Documentation/devicetree/bindings/timer/renesas,sh-cmt-timer.txt new file +8 -0
@@ -0,0 +1,21 @@
+* Renesas SH Mobile Compare Match Timer
+
+Required properties:
+- compatible : Should be "renesas,cmt"
+- reg : Address and length of the register set for the device
+- interrupts : Timer IRQ
+- renesas,timer-device-id : The ID of the timer device
+- renesas,timer-channel-id : The channel ID of the timer device

I'm not familiar with this hardware. Could you give me a basic idea of how it's structured?

Quoted 2 lines

Does the device have a single timer that this describes, or does the device have multiple timers, and this selects which one to use?

Quoted 9 lines

The CMT timer is built into various SoCs from Renesas, often multiple instances of it. Each timer instance has 6 channels. You can select if they are driven by the main clock or the real time clock (the SoCs include an SH core for data processing) and can use different subscale modes. It features free running mode and can fire events. The clock has up to 48 bits to count. It has a DMA channel and wires enabling the device to be a wakeup source. Hmm don't know what to add more. It's a timer - you can check for overflows, read the counter and so on.

Quoted 5 lines

Unless I've misunderstood, couldn't Linux choose one timer to use completely arbitrarily?

Why do we need to describe which one to use?

Many of these CPUs have multiple timer blocks (CMT, TMU, MTU2) that sit in different power domains, so there is often a desire to prioritize one over another, as it will determine which sleep states are available to us. Beyond that, many of these channels have special behavioural constraints (only certain channels being useable by the PWM block for example, as well as mux issues (certain channels will be tied up with functions we care less about than others).

We have traditionally registered all of the available timer channels as platform devices, and then left it to the clocksource/event rating to determine which to use for what. This was based on the assumption that later on we would do some work on the clockevents layer to dynamically allocate the other channels that weren't being used for anything, but this never happened, and most users managed to get away with simply adopting hrtimers instead.

Quoted 5 lines

Perhaps the rating is not a detail we want to expose, but we are still going to need some facility for defining whether a given timer is suitable for a clocksource/clockevent or not. We also can't really encode the rating information in the driver itself, as the placement of the block is wholly SoC dependent.

Thanks for the explanation. Given this situation I wonder if we can add the optional properties:
renesas,source-quality : The viability to use this device as a free running clock
renesas,event-quality : Th viability to use this device as an event generator

Maybe let it run from 0-10 or from 0-255 and explain the situation in the bindings.
I mean the environmental circumstances of the device are not really a property of the device itself, but it doesn't need much imagination to see it like that.

Thanks,

Bastian

#9 lore raw

PATCH 2/3 ARM: mach-shmobile: sh73a0: Add CMT DT name to clock list

In reply to Bastian Hecht

This adds temporarily the alternative device name to the clock list that is used when booting via Device Tree setup.

---
 arch/arm/mach-shmobile/clock-sh73a0.c |    1 +
 1 file changed, 1 insertion(+)
arch/arm/mach-shmobile/clock-sh73a0.c +1 -0
index 71843dd..4aed113 100644
@@ -575,6 +575,7 @@ static struct clk_lookup lookups[] = {
 	CLKDEV_DEV_ID("sh-sci.4", &mstp_clks[MSTP200]), /* SCIFA4 */
 	CLKDEV_DEV_ID("sh-sci.6", &mstp_clks[MSTP331]), /* SCIFA6 */
 	CLKDEV_DEV_ID("sh_cmt.10", &mstp_clks[MSTP329]), /* CMT10 */
+	CLKDEV_DEV_ID("e6138010.cmt", &mstp_clks[MSTP329]), /* CMT10 */
 	CLKDEV_DEV_ID("sh_fsi2", &mstp_clks[MSTP328]), /* FSI */
 	CLKDEV_DEV_ID("sh_irda.0", &mstp_clks[MSTP325]), /* IrDA */
 	CLKDEV_DEV_ID("i2c-sh_mobile.1", &mstp_clks[MSTP323]), /* I2C1 */
Signature · 1 line
1.7.9.5
#10 lore raw

PATCH 3/3 ARM: mach-shmobile: sh73a0: Setup the timer device CMT10 using DT

In reply to Bastian Hecht

We can now use the Device Tree for bringing up our timer device CMT10. We move it out of the DT devices list into the early_devices list by the non-reference board code. And we add the device to the kzm9g-reference .dts file.

---
I wonder if this should go to arch/arm/boot/dts/sh73a0.dtsi.

The moving to the early_devices is a bit hackish - I will cook up a better patch when I know into which .dts(i) file things go.

 arch/arm/boot/dts/sh73a0-kzm9g-reference.dts |   11 +++++++++++
 arch/arm/mach-shmobile/setup-sh73a0.c        |    2 +-
 2 files changed, 12 insertions(+), 1 deletion(-)
arch/arm/boot/dts/sh73a0-kzm9g-reference.dts +11 -0
index 7fad4b9..fef12f0 100644
@@ -69,6 +69,17 @@
 		toshiba,mmc-wrprotect-disable;
 		toshiba,mmc-has-idle-wait;
 	};
+
+	cmt@0xe6138010 {
+		compatible = "renesas,cmt";
+		reg = <0xe6138010 0xc>;
+		interrupt-parent = <&gic>;
+		interrupts = <0 65 0x4>;
+		renesas,timer-device-id = <1>;
+		renesas,timer-channel-id = <0>;
+		renesas,clocksource-rating = <125>;
+		renesas,clockevent-rating = <125>;
+	};
 };
 
 &mmcif {
arch/arm/mach-shmobile/setup-sh73a0.c +1 -1
index 6259e07..b7fdec0 100644
@@ -913,10 +913,10 @@ static struct platform_device *sh73a0_devices_dt[] __initdata = {
 	&scif6_device,
 	&scif7_device,
 	&scif8_device,
-	&cmt10_device,
 };
 
 static struct platform_device *sh73a0_early_devices[] __initdata = {
+	&cmt10_device,
 	&tmu00_device,
 	&tmu01_device,
 };
Signature · 1 line
1.7.9.5
#11 lore raw

PATCH 3/3 ARM: mach-shmobile: sh73a0: Setup the timer device CMT10 using DT

Hello.

Quoted 21 lines · Bastian Hecht

We can now use the Device Tree for bringing up our timer device CMT10. We move it out of the DT devices list into the early_devices list by the non-reference board code. And we add the device to the kzm9g-reference .dts file.

Not-yet-signed-off-by Bastian Hecht<[email protected]>

---
I wonder if this should go to arch/arm/boot/dts/sh73a0.dtsi.

The moving to the early_devices is a bit hackish - I will cook up a better patch when I know into which .dts(i) file things go.

  arch/arm/boot/dts/sh73a0-kzm9g-reference.dts |   11 +++++++++++
  arch/arm/mach-shmobile/setup-sh73a0.c        |    2 +-
  2 files changed, 12 insertions(+), 1 deletion(-)
arch/arm/boot/dts/sh73a0-kzm9g-reference.dts +0 -0
index 7fad4b9..fef12f0 100644

arch/arm/boot/dts/sh73a0-kzm9g-reference.dts line 74

Quoted 6 lines
arch/arm/boot/dts/sh73a0-kzm9g-reference.dts +2 -0
@@ -69,6 +69,17 @@
  		toshiba,mmc-wrprotect-disable;
  		toshiba,mmc-has-idle-wait;
  	};
+
+	cmt@0xe6138010 {
 
     "0x" shouldn't be there, remove it please.

WBR, Sergei

#12 lore raw

PATCH 3/3 ARM: mach-shmobile: sh73a0: Setup the timer device CMT10 using DT

Hi,

2013/3/1 Sergei Shtylyov <[email protected]>:

Quoted 37 lines

Hello.

Quoted 22 lines · Bastian Hecht

We can now use the Device Tree for bringing up our timer device CMT10. We move it out of the DT devices list into the early_devices list by the non-reference board code. And we add the device to the kzm9g-reference .dts file.

Not-yet-signed-off-by Bastian Hecht<[email protected]>

---
I wonder if this should go to arch/arm/boot/dts/sh73a0.dtsi.

The moving to the early_devices is a bit hackish - I will cook up a better patch when I know into which .dts(i) file things go.

  arch/arm/boot/dts/sh73a0-kzm9g-reference.dts |   11 +++++++++++
  arch/arm/mach-shmobile/setup-sh73a0.c        |    2 +-
  2 files changed, 12 insertions(+), 1 deletion(-)
arch/arm/boot/dts/sh73a0-kzm9g-reference.dts +0 -0
index 7fad4b9..fef12f0 100644

arch/arm/boot/dts/sh73a0-kzm9g-reference.dts line 74

Quoted 6 lines
arch/arm/boot/dts/sh73a0-kzm9g-reference.dts +2 -0
@@ -69,6 +69,17 @@
                toshiba,mmc-wrprotect-disable;
                toshiba,mmc-has-idle-wait;
        };
+
+       cmt at 0xe6138010 {
 
    "0x" shouldn't be there, remove it please.
Quoted 2 lines

WBR, Sergei

Thanks for pointing out this one. Fixed it in another patchset as well.

Bastian