diff mbox

[v2,02/16] arm64: dts: add sp804 timer node for Hi6220

Message ID 1459589383-16914-3-git-send-email-guodong.xu@linaro.org (mailing list archive)
State New, archived
Headers show

Commit Message

Guodong Xu April 2, 2016, 9:29 a.m. UTC
From: Leo Yan <leo.yan@linaro.org>

Add sp804 timer for hi6220, so it can be used as broadcast timer.

Signed-off-by: Leo Yan <leo.yan@linaro.org>
Signed-off-by: Wei Xu <xuwei5@hisilicon.com>
---
 arch/arm64/boot/dts/hisilicon/hi6220.dtsi | 9 +++++++++
 1 file changed, 9 insertions(+)

Comments

Linus Walleij April 3, 2016, 7:23 p.m. UTC | #1
On Sat, Apr 2, 2016 at 11:29 AM, Guodong Xu <guodong.xu@linaro.org> wrote:

> From: Leo Yan <leo.yan@linaro.org>
>
> Add sp804 timer for hi6220, so it can be used as broadcast timer.
>
> Signed-off-by: Leo Yan <leo.yan@linaro.org>
> Signed-off-by: Wei Xu <xuwei5@hisilicon.com>
> ---
>  arch/arm64/boot/dts/hisilicon/hi6220.dtsi | 9 +++++++++
>  1 file changed, 9 insertions(+)
>
> diff --git a/arch/arm64/boot/dts/hisilicon/hi6220.dtsi b/arch/arm64/boot/dts/hisilicon/hi6220.dtsi
> index ad1f1eb..82c4756 100644
> --- a/arch/arm64/boot/dts/hisilicon/hi6220.dtsi
> +++ b/arch/arm64/boot/dts/hisilicon/hi6220.dtsi
> @@ -209,5 +209,14 @@
>                         clock-names = "uartclk", "apb_pclk";
>                         status = "disabled";
>                 };
> +
> +               dual_timer0: dual_timer@f8008000 {
> +                       compatible = "arm,sp804", "arm,primecell";
> +                       reg = <0x0 0xf8008000 0x0 0x1000>;
> +                       interrupts = <GIC_SPI 14 IRQ_TYPE_LEVEL_HIGH>,
> +                                    <GIC_SPI 15 IRQ_TYPE_LEVEL_HIGH>;
> +                       clocks = <&ao_ctrl 27>;
> +                       clock-names = "apb_pclk";

How can this work? You only give the apb_pclk for clocking the
bus to the timer.

Most platforms using this driver has something like this:

                timer01: timer@10104000 {
                        compatible = "arm,sp804", "arm,primecell";
                        reg = <0x10104000 0x1000>;
                        interrupt-parent = <&intc_dc1176>;
                        interrupts = <0 8 IRQ_TYPE_LEVEL_HIGH>, <0 9
IRQ_TYPE_LEVEL_HIGH>;
                        clocks = <&timclk>, <&timclk>, <&pclk>;
                        clock-names = "timer1", "timer2", "apb_pclk";
                };

It then reads the two clocks in the beginning of clocks() to
determine the frequency of each timer.

By chance the code in the driver will allow just one clock and
will then assume that both the bus to the timer and the timer
itself is clocked from the same clock. But I highly doubt that this
is the case.

Please verify what clocks actually goes into this timer, it should
nominally be three of them.

Yours,
Linus Walleij
Leo Yan April 4, 2016, 1:43 a.m. UTC | #2
Hi Linus,

Thanks for review.

On Sun, Apr 03, 2016 at 09:23:42PM +0200, Linus Walleij wrote:
> On Sat, Apr 2, 2016 at 11:29 AM, Guodong Xu <guodong.xu@linaro.org> wrote:
> 
> > From: Leo Yan <leo.yan@linaro.org>
> >
> > Add sp804 timer for hi6220, so it can be used as broadcast timer.
> >
> > Signed-off-by: Leo Yan <leo.yan@linaro.org>
> > Signed-off-by: Wei Xu <xuwei5@hisilicon.com>
> > ---
> >  arch/arm64/boot/dts/hisilicon/hi6220.dtsi | 9 +++++++++
> >  1 file changed, 9 insertions(+)
> >
> > diff --git a/arch/arm64/boot/dts/hisilicon/hi6220.dtsi b/arch/arm64/boot/dts/hisilicon/hi6220.dtsi
> > index ad1f1eb..82c4756 100644
> > --- a/arch/arm64/boot/dts/hisilicon/hi6220.dtsi
> > +++ b/arch/arm64/boot/dts/hisilicon/hi6220.dtsi
> > @@ -209,5 +209,14 @@
> >                         clock-names = "uartclk", "apb_pclk";
> >                         status = "disabled";
> >                 };
> > +
> > +               dual_timer0: dual_timer@f8008000 {
> > +                       compatible = "arm,sp804", "arm,primecell";
> > +                       reg = <0x0 0xf8008000 0x0 0x1000>;
> > +                       interrupts = <GIC_SPI 14 IRQ_TYPE_LEVEL_HIGH>,
> > +                                    <GIC_SPI 15 IRQ_TYPE_LEVEL_HIGH>;
> > +                       clocks = <&ao_ctrl 27>;
> > +                       clock-names = "apb_pclk";
> 
> How can this work? You only give the apb_pclk for clocking the
> bus to the timer.
> 
> Most platforms using this driver has something like this:
> 
>                 timer01: timer@10104000 {
>                         compatible = "arm,sp804", "arm,primecell";
>                         reg = <0x10104000 0x1000>;
>                         interrupt-parent = <&intc_dc1176>;
>                         interrupts = <0 8 IRQ_TYPE_LEVEL_HIGH>, <0 9
> IRQ_TYPE_LEVEL_HIGH>;
>                         clocks = <&timclk>, <&timclk>, <&pclk>;
>                         clock-names = "timer1", "timer2", "apb_pclk";
>                 };
> 
> It then reads the two clocks in the beginning of clocks() to
> determine the frequency of each timer.
> 
> By chance the code in the driver will allow just one clock and
> will then assume that both the bus to the timer and the timer
> itself is clocked from the same clock. But I highly doubt that this
> is the case.

This patch has been sent out for review previously [1]; So I refered
other platforms and changed to only enable apb bus clock due I have
not found timer enabling bits in Hi6220's spec.

> Please verify what clocks actually goes into this timer, it should
> nominally be three of them.

Yes, it can work well even I don't enable any timer clock. I will
check furthermore in spec to see if can bind with formal way as your
suggestion.

[1] http://archive.arm.linux.org.uk/lurker/message/20151009.131015.687ed525.en.html

Thanks,
Leo Yan
Rob Herring (Arm) April 4, 2016, 5:16 a.m. UTC | #3
On Sun, Apr 03, 2016 at 09:23:42PM +0200, Linus Walleij wrote:
> On Sat, Apr 2, 2016 at 11:29 AM, Guodong Xu <guodong.xu@linaro.org> wrote:
> 
> > From: Leo Yan <leo.yan@linaro.org>
> >
> > Add sp804 timer for hi6220, so it can be used as broadcast timer.
> >
> > Signed-off-by: Leo Yan <leo.yan@linaro.org>
> > Signed-off-by: Wei Xu <xuwei5@hisilicon.com>
> > ---
> >  arch/arm64/boot/dts/hisilicon/hi6220.dtsi | 9 +++++++++
> >  1 file changed, 9 insertions(+)
> >
> > diff --git a/arch/arm64/boot/dts/hisilicon/hi6220.dtsi b/arch/arm64/boot/dts/hisilicon/hi6220.dtsi
> > index ad1f1eb..82c4756 100644
> > --- a/arch/arm64/boot/dts/hisilicon/hi6220.dtsi
> > +++ b/arch/arm64/boot/dts/hisilicon/hi6220.dtsi
> > @@ -209,5 +209,14 @@
> >                         clock-names = "uartclk", "apb_pclk";
> >                         status = "disabled";
> >                 };
> > +
> > +               dual_timer0: dual_timer@f8008000 {

timer@...

> > +                       compatible = "arm,sp804", "arm,primecell";
> > +                       reg = <0x0 0xf8008000 0x0 0x1000>;
> > +                       interrupts = <GIC_SPI 14 IRQ_TYPE_LEVEL_HIGH>,
> > +                                    <GIC_SPI 15 IRQ_TYPE_LEVEL_HIGH>;
> > +                       clocks = <&ao_ctrl 27>;
> > +                       clock-names = "apb_pclk";
> 
> How can this work? You only give the apb_pclk for clocking the
> bus to the timer.
> 
> Most platforms using this driver has something like this:
> 
>                 timer01: timer@10104000 {
>                         compatible = "arm,sp804", "arm,primecell";
>                         reg = <0x10104000 0x1000>;
>                         interrupt-parent = <&intc_dc1176>;
>                         interrupts = <0 8 IRQ_TYPE_LEVEL_HIGH>, <0 9
> IRQ_TYPE_LEVEL_HIGH>;
>                         clocks = <&timclk>, <&timclk>, <&pclk>;
>                         clock-names = "timer1", "timer2", "apb_pclk";
>                 };
> 
> It then reads the two clocks in the beginning of clocks() to
> determine the frequency of each timer.
> 
> By chance the code in the driver will allow just one clock and
> will then assume that both the bus to the timer and the timer
> itself is clocked from the same clock. But I highly doubt that this
> is the case.

I believe 1 clock was allowed for early bindings, but yes it should be 
3.

Rob
Linus Walleij April 4, 2016, 11:21 a.m. UTC | #4
On Mon, Apr 4, 2016 at 3:43 AM, Leo Yan <leo.yan@linaro.org> wrote:
> On Sun, Apr 03, 2016 at 09:23:42PM +0200, Linus Walleij wrote:
>> On Sat, Apr 2, 2016 at 11:29 AM, Guodong Xu <guodong.xu@linaro.org> wrote:

>> By chance the code in the driver will allow just one clock and
>> will then assume that both the bus to the timer and the timer
>> itself is clocked from the same clock. But I highly doubt that this
>> is the case.
>
> This patch has been sent out for review previously [1]; So I refered
> other platforms and changed to only enable apb bus clock due I have
> not found timer enabling bits in Hi6220's spec.

This is not about enabling/disabling the clock(s) to the timer.
It doesn't matter if these clocks are always on.

It is about determining the *frequency* of the timers.

It is vital that the timer driver get the right frequency of the clock
to the block from the clock implementation, and I do not think
it is the same as the "apb_pclk".

The thing is that of course "any frequency" will
work but what you will notice is that the timer runs very
weirdly compared to wall-clock time unless the right clock
yielding the right frequency has been specified here.

Yours,
Linus Walleij
Leo Yan April 4, 2016, 1:42 p.m. UTC | #5
On Mon, Apr 04, 2016 at 01:21:00PM +0200, Linus Walleij wrote:
> On Mon, Apr 4, 2016 at 3:43 AM, Leo Yan <leo.yan@linaro.org> wrote:
> > On Sun, Apr 03, 2016 at 09:23:42PM +0200, Linus Walleij wrote:
> >> On Sat, Apr 2, 2016 at 11:29 AM, Guodong Xu <guodong.xu@linaro.org> wrote:
> 
> >> By chance the code in the driver will allow just one clock and
> >> will then assume that both the bus to the timer and the timer
> >> itself is clocked from the same clock. But I highly doubt that this
> >> is the case.
> >
> > This patch has been sent out for review previously [1]; So I refered
> > other platforms and changed to only enable apb bus clock due I have
> > not found timer enabling bits in Hi6220's spec.
> 
> This is not about enabling/disabling the clock(s) to the timer.
> It doesn't matter if these clocks are always on.
> 
> It is about determining the *frequency* of the timers.
> 
> It is vital that the timer driver get the right frequency of the clock
> to the block from the clock implementation, and I do not think
> it is the same as the "apb_pclk".

Thanks for reminding. Fortunately, apb_pclk and timer's clock are
same, all of them's rate are 19.2MHz.

> The thing is that of course "any frequency" will
> work but what you will notice is that the timer runs very
> weirdly compared to wall-clock time unless the right clock
> yielding the right frequency has been specified here.

Compared sp804 timer counter with wall-clock (which is using ARM's
arch timer), and confirmed that sp804 timer's counter register is
decreasing with rate 19.2MHz.

Also have checked Hi6220's spec, there have no timer's dediated clock
enabling bits. This is the reason before I only registered one clock.
So according to you and Rob's comments, how about change as below?

		dual_timer0: timer@f8008000 {
			compatible = "arm,sp804", "arm,primecell";
			reg = <0x0 0xf8008000 0x0 0x1000>;
			interrupts = <GIC_SPI 14 IRQ_TYPE_LEVEL_HIGH>,
				     <GIC_SPI 15 IRQ_TYPE_LEVEL_HIGH>;
			clocks = <&ao_ctrl HI6220_TIMER0_PCLK>,
				 <&ao_ctrl HI6220_TIMER0_PCLK>,
				 <&ao_ctrl HI6220_TIMER0_PCLK>;
			clock-names = "apb_pclk", "apb_pclk", "apb_pclk";
		};

Thanks,
Leo Yan
Linus Walleij April 4, 2016, 1:53 p.m. UTC | #6
On Mon, Apr 4, 2016 at 3:42 PM, Leo Yan <leo.yan@linaro.org> wrote:
> On Mon, Apr 04, 2016 at 01:21:00PM +0200, Linus Walleij wrote:

>> This is not about enabling/disabling the clock(s) to the timer.
>> It doesn't matter if these clocks are always on.
>>
>> It is about determining the *frequency* of the timers.
>>
>> It is vital that the timer driver get the right frequency of the clock
>> to the block from the clock implementation, and I do not think
>> it is the same as the "apb_pclk".
>
> Thanks for reminding. Fortunately, apb_pclk and timer's clock are
> same, all of them's rate are 19.2MHz.

OK I was wrong, worrying about nothing.

>> The thing is that of course "any frequency" will
>> work but what you will notice is that the timer runs very
>> weirdly compared to wall-clock time unless the right clock
>> yielding the right frequency has been specified here.
>
> Compared sp804 timer counter with wall-clock (which is using ARM's
> arch timer), and confirmed that sp804 timer's counter register is
> decreasing with rate 19.2MHz.

Thanks, awesome.

> Also have checked Hi6220's spec, there have no timer's dediated clock
> enabling bits. This is the reason before I only registered one clock.
> So according to you and Rob's comments, how about change as below?
>
>                 dual_timer0: timer@f8008000 {
>                         compatible = "arm,sp804", "arm,primecell";
>                         reg = <0x0 0xf8008000 0x0 0x1000>;
>                         interrupts = <GIC_SPI 14 IRQ_TYPE_LEVEL_HIGH>,
>                                      <GIC_SPI 15 IRQ_TYPE_LEVEL_HIGH>;
>                         clocks = <&ao_ctrl HI6220_TIMER0_PCLK>,
>                                  <&ao_ctrl HI6220_TIMER0_PCLK>,
>                                  <&ao_ctrl HI6220_TIMER0_PCLK>;
>                         clock-names = "apb_pclk", "apb_pclk", "apb_pclk";

This works for me, though I think only the last name
matters so I would name the first two "timer1" and "timer2".

Yours,
Linus Walleij
Leo Yan April 4, 2016, 2:03 p.m. UTC | #7
On Mon, Apr 04, 2016 at 03:53:47PM +0200, Linus Walleij wrote:
> On Mon, Apr 4, 2016 at 3:42 PM, Leo Yan <leo.yan@linaro.org> wrote:
> > On Mon, Apr 04, 2016 at 01:21:00PM +0200, Linus Walleij wrote:

[...]

> > Also have checked Hi6220's spec, there have no timer's dediated clock
> > enabling bits. This is the reason before I only registered one clock.
> > So according to you and Rob's comments, how about change as below?
> >
> >                 dual_timer0: timer@f8008000 {
> >                         compatible = "arm,sp804", "arm,primecell";
> >                         reg = <0x0 0xf8008000 0x0 0x1000>;
> >                         interrupts = <GIC_SPI 14 IRQ_TYPE_LEVEL_HIGH>,
> >                                      <GIC_SPI 15 IRQ_TYPE_LEVEL_HIGH>;
> >                         clocks = <&ao_ctrl HI6220_TIMER0_PCLK>,
> >                                  <&ao_ctrl HI6220_TIMER0_PCLK>,
> >                                  <&ao_ctrl HI6220_TIMER0_PCLK>;
> >                         clock-names = "apb_pclk", "apb_pclk", "apb_pclk";
> 
> This works for me, though I think only the last name
> matters so I would name the first two "timer1" and "timer2".

Will fix to "timer1" and "timer2" and sent out new patch.

Thanks,
Leo Yan
diff mbox

Patch

diff --git a/arch/arm64/boot/dts/hisilicon/hi6220.dtsi b/arch/arm64/boot/dts/hisilicon/hi6220.dtsi
index ad1f1eb..82c4756 100644
--- a/arch/arm64/boot/dts/hisilicon/hi6220.dtsi
+++ b/arch/arm64/boot/dts/hisilicon/hi6220.dtsi
@@ -209,5 +209,14 @@ 
 			clock-names = "uartclk", "apb_pclk";
 			status = "disabled";
 		};
+
+		dual_timer0: dual_timer@f8008000 {
+			compatible = "arm,sp804", "arm,primecell";
+			reg = <0x0 0xf8008000 0x0 0x1000>;
+			interrupts = <GIC_SPI 14 IRQ_TYPE_LEVEL_HIGH>,
+				     <GIC_SPI 15 IRQ_TYPE_LEVEL_HIGH>;
+			clocks = <&ao_ctrl 27>;
+			clock-names = "apb_pclk";
+		};
 	};
 };