[PATCH] arm64: dts: mt8173: add clock_null

[PATCH] arm64: dts: mt8173: add clock_null

[Eddie Huang]

Add clk_null, which represents clocks that can not / need not
controlled by software.
There are many clocks' parent set to clk_null.

Signed-off-by: James Liao <jamesjj.liao@mediatek.com>
Signed-off-by: Eddie Huang <eddie.huang@mediatek.com>
---
Base on 4.1-rc1

Change-Id: I4db9b40d07e28f54f7bae9b676316cbd6a962124
---
 arch/arm64/boot/dts/mediatek/mt8173.dtsi | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/arch/arm64/boot/dts/mediatek/mt8173.dtsi b/arch/arm64/boot/dts/mediatek/mt8173.dtsi
index 924fdb6..4798f44 100644
--- a/arch/arm64/boot/dts/mediatek/mt8173.dtsi
+++ b/arch/arm64/boot/dts/mediatek/mt8173.dtsi
@@ -81,6 +81,12 @@
 		cpu_on	      = <0x84000003>;
 	};
 
+	clk_null: clk_null {
+		compatible = "fixed-clock";
+		clock-frequency = <0>;
+		#clock-cells = <0>;
+	};
+
 	uart_clk: dummy26m {
 		compatible = "fixed-clock";
 		clock-frequency = <26000000>;
-- 
1.8.1.1.dirty

[PATCH] arm64: dts: mt8173: add clock_null

[Heiko Stuebner]

Am Donnerstag, 18. Juni 2015, 13:29:11 schrieb Eddie Huang:
> Add clk_null, which represents clocks that can not / need not
> controlled by software.
> There are many clocks' parent set to clk_null.

Devicetree is supposed to describe hardware, and ideally not what software 
does with it. If the clock simply cannot be controlled by software, it will 
still have a rate and I think it should probably be modelled - similarly we 
sometimes have fixed regulators that also are not software controllable.


While it might be ok to define dummy clocks as a temporary stopgap, these 
should definitly be marked as such. This clk_null at least sounds like there is 
no plan to replace this with a real solution at some point.

And of course a bit of context would be cool, to know which type of clocks 
this actually replaces.


Heiko

> Signed-off-by: James Liao <jamesjj.liao@mediatek.com>
> Signed-off-by: Eddie Huang <eddie.huang@mediatek.com>
> ---
> Base on 4.1-rc1
> 
> Change-Id: I4db9b40d07e28f54f7bae9b676316cbd6a962124
> ---
>  arch/arm64/boot/dts/mediatek/mt8173.dtsi | 6 ++++++
>  1 file changed, 6 insertions(+)
> 
> diff --git a/arch/arm64/boot/dts/mediatek/mt8173.dtsi
> b/arch/arm64/boot/dts/mediatek/mt8173.dtsi index 924fdb6..4798f44 100644
> --- a/arch/arm64/boot/dts/mediatek/mt8173.dtsi
> +++ b/arch/arm64/boot/dts/mediatek/mt8173.dtsi
> @@ -81,6 +81,12 @@
>  		cpu_on	      = <0x84000003>;
>  	};
> 
> +	clk_null: clk_null {
> +		compatible = "fixed-clock";
> +		clock-frequency = <0>;
> +		#clock-cells = <0>;
> +	};
> +
>  	uart_clk: dummy26m {
>  		compatible = "fixed-clock";
>  		clock-frequency = <26000000>;

[PATCH] arm64: dts: mt8173: add clock_null

[Heiko Stuebner]

Am Donnerstag, 18. Juni 2015, 18:15:03 schrieb Heiko Stuebner:
> Am Donnerstag, 18. Juni 2015, 13:29:11 schrieb Eddie Huang:
> > Add clk_null, which represents clocks that can not / need not
> > controlled by software.
> > There are many clocks' parent set to clk_null.
> 
> Devicetree is supposed to describe hardware, and ideally not what software
> does with it. If the clock simply cannot be controlled by software, it will
> still have a rate and I think it should probably be modelled - similarly we
> sometimes have fixed regulators that also are not software controllable.
> 
> 
> While it might be ok to define dummy clocks as a temporary stopgap, these
> should definitly be marked as such. This clk_null at least sounds like there
> is no plan to replace this with a real solution at some point.
> 
> And of course a bit of context would be cool, to know which type of clocks
> this actually replaces.

After looking a bit more into this, I'm feel that the clk_null approach is 
wrong.

For one, even if the clk_null stuff would be ok, the binding doc of the clock 
controller does not describe the need of this specific clock at all.


The other more important point, looking at clk-mt8173 I see at least these 
clocks being set as children of "clk_null":

static const struct mtk_fixed_factor root_clk_alias[] __initconst = {
	FACTOR(CLK_TOP_CLKPH_MCK_O, "clkph_mck_o", "clk_null", 1, 1),
	FACTOR(CLK_TOP_DPI, "dpi_ck", "clk_null", 1, 1),
	FACTOR(CLK_TOP_USB_SYSPLL_125M, "usb_syspll_125m", "clk_null", 1, 1),
	FACTOR(CLK_TOP_HDMITX_DIG_CTS, "hdmitx_dig_cts", "clk_null", 1, 1),
};


These look more like they are fed from some external source to the clock 
controller? I did ask Matthias but he also couldn't find anything describing 
where these clocks actually come from.


Heiko

[PATCH] arm64: dts: mt8173: add clock_null

[James Liao]

Hi Heiko,

On Fri, 2015-06-19 at 13:36 +0200, Heiko Stuebner wrote:
> Am Donnerstag, 18. Juni 2015, 18:15:03 schrieb Heiko Stuebner:
> > Am Donnerstag, 18. Juni 2015, 13:29:11 schrieb Eddie Huang:
> > > Add clk_null, which represents clocks that can not / need not
> > > controlled by software.
> > > There are many clocks' parent set to clk_null.
> > 
> > Devicetree is supposed to describe hardware, and ideally not what software
> > does with it. If the clock simply cannot be controlled by software, it will
> > still have a rate and I think it should probably be modelled - similarly we
> > sometimes have fixed regulators that also are not software controllable.
> > 
> > 
> > While it might be ok to define dummy clocks as a temporary stopgap, these
> > should definitly be marked as such. This clk_null at least sounds like there
> > is no plan to replace this with a real solution at some point.
> > 
> > And of course a bit of context would be cool, to know which type of clocks
> > this actually replaces.
> 
> After looking a bit more into this, I'm feel that the clk_null approach is 
> wrong.
> 
> For one, even if the clk_null stuff would be ok, the binding doc of the clock 
> controller does not describe the need of this specific clock at all.
> 
> 
> The other more important point, looking at clk-mt8173 I see at least these 
> clocks being set as children of "clk_null":
> 
> static const struct mtk_fixed_factor root_clk_alias[] __initconst = {
> 	FACTOR(CLK_TOP_CLKPH_MCK_O, "clkph_mck_o", "clk_null", 1, 1),
> 	FACTOR(CLK_TOP_DPI, "dpi_ck", "clk_null", 1, 1),
> 	FACTOR(CLK_TOP_USB_SYSPLL_125M, "usb_syspll_125m", "clk_null", 1, 1),
> 	FACTOR(CLK_TOP_HDMITX_DIG_CTS, "hdmitx_dig_cts", "clk_null", 1, 1),
> };
> 
> 
> These look more like they are fed from some external source to the clock 
> controller? I did ask Matthias but he also couldn't find anything describing 
> where these clocks actually come from.

Some clocks such as clkph_mck_o, we don't really care where they come
from and what frequencies are. We model these clocks just because they
or their derived clocks can be the source of topckgen muxes. Is there a
better way to model "don't care" clocks?


Best regards,

James

[PATCH] arm64: dts: mt8173: add clock_null

[Heiko Stübner]

Hi James,

Am Montag, 22. Juni 2015, 11:38:37 schrieb James Liao:
> On Fri, 2015-06-19 at 13:36 +0200, Heiko Stuebner wrote:
> > Am Donnerstag, 18. Juni 2015, 18:15:03 schrieb Heiko Stuebner:
> > > Am Donnerstag, 18. Juni 2015, 13:29:11 schrieb Eddie Huang:
> > > > Add clk_null, which represents clocks that can not / need not
> > > > controlled by software.
> > > > There are many clocks' parent set to clk_null.
> > > 
> > > Devicetree is supposed to describe hardware, and ideally not what
> > > software
> > > does with it. If the clock simply cannot be controlled by software, it
> > > will
> > > still have a rate and I think it should probably be modelled - similarly
> > > we
> > > sometimes have fixed regulators that also are not software controllable.
> > > 
> > > 
> > > While it might be ok to define dummy clocks as a temporary stopgap,
> > > these
> > > should definitly be marked as such. This clk_null at least sounds like
> > > there is no plan to replace this with a real solution at some point.
> > > 
> > > And of course a bit of context would be cool, to know which type of
> > > clocks
> > > this actually replaces.
> > 
> > After looking a bit more into this, I'm feel that the clk_null approach is
> > wrong.
> > 
> > For one, even if the clk_null stuff would be ok, the binding doc of the
> > clock controller does not describe the need of this specific clock at
> > all.
> > 
> > 
> > The other more important point, looking at clk-mt8173 I see at least these
> > clocks being set as children of "clk_null":
> > 
> > static const struct mtk_fixed_factor root_clk_alias[] __initconst = {
> > 
> > 	FACTOR(CLK_TOP_CLKPH_MCK_O, "clkph_mck_o", "clk_null", 1, 1),
> > 	FACTOR(CLK_TOP_DPI, "dpi_ck", "clk_null", 1, 1),
> > 	FACTOR(CLK_TOP_USB_SYSPLL_125M, "usb_syspll_125m", "clk_null", 1, 1),
> > 	FACTOR(CLK_TOP_HDMITX_DIG_CTS, "hdmitx_dig_cts", "clk_null", 1, 1),
> > 
> > };
> > 
> > 
> > These look more like they are fed from some external source to the clock
> > controller? I did ask Matthias but he also couldn't find anything
> > describing where these clocks actually come from.
> 
> Some clocks such as clkph_mck_o, we don't really care where they come
> from and what frequencies are. We model these clocks just because they
> or their derived clocks can be the source of topckgen muxes. Is there a
> better way to model "don't care" clocks?

There are two different concepts at work here. You might not care in your 
kernel driver implementation _at the moment_ where the clocks come from; but 
the devicetree is supposed to model how the hardware is structured and not 
contain implementation specific details.

So the clock tree should be modeled according to how the hardware is layed out 
not how you want to use the clocks at the moment :-) .

It would it any case be good, if you could describe where these clocks come 
from in the hardware, so we can find the best solution on how to model those.


Heiko

[PATCH] arm64: dts: mt8173: add clock_null

[James Liao]

Hi Heiko,

On Mon, 2015-06-22 at 14:53 +0200, Heiko St?bner wrote:
> Hi James,
> 
> Am Montag, 22. Juni 2015, 11:38:37 schrieb James Liao:
> > On Fri, 2015-06-19 at 13:36 +0200, Heiko Stuebner wrote:
> > 
> > Some clocks such as clkph_mck_o, we don't really care where they come
> > from and what frequencies are. We model these clocks just because they
> > or their derived clocks can be the source of topckgen muxes. Is there a
> > better way to model "don't care" clocks?
> 
> There are two different concepts at work here. You might not care in your 
> kernel driver implementation _at the moment_ where the clocks come from; but 
> the devicetree is supposed to model how the hardware is structured and not 
> contain implementation specific details.

If we model "don't care" clocks inside the driver (i.e. create clk_null
in clock driver), then we don't need to model the dummy clock in device.
Is it an acceptable way?

> So the clock tree should be modeled according to how the hardware is layed out 
> not how you want to use the clocks at the moment :-) .
> 
> It would it any case be good, if you could describe where these clocks come 
> from in the hardware, so we can find the best solution on how to model those.

In fact, I don't know where these clocks come from at all, especially
when they come from outside of SoC. Besides, some clocks don't need to
model in CCF, but they can be the source of clocks that controlled by
CCF.

I don't think ALL clocks on a SoC need to be handled in CCF, so there
should be some clocks don't have a "real" or "correct" parent. In
current implementation I use a dummy clock (clk_null) to be the unreal
parent.

Do you think we should model as more clocks as we can in CCF even we
don't need them? If no, how do we handle those clocks which are not
handled in CCF but can be parent of CCF clocks?


Best regards,

James

[PATCH] arm64: dts: mt8173: add clock_null

[Heiko Stübner]

Hi James,

Am Mittwoch, 24. Juni 2015, 15:54:15 schrieb James Liao:
> On Mon, 2015-06-22 at 14:53 +0200, Heiko St?bner wrote:
> > Hi James,
> > 
> > Am Montag, 22. Juni 2015, 11:38:37 schrieb James Liao:
> > > On Fri, 2015-06-19 at 13:36 +0200, Heiko Stuebner wrote:
> > > 
> > > Some clocks such as clkph_mck_o, we don't really care where they come
> > > from and what frequencies are. We model these clocks just because they
> > > or their derived clocks can be the source of topckgen muxes. Is there a
> > > better way to model "don't care" clocks?
> > 
> > There are two different concepts at work here. You might not care in your
> > kernel driver implementation _at the moment_ where the clocks come from;
> > but the devicetree is supposed to model how the hardware is structured
> > and not contain implementation specific details.
> 
> If we model "don't care" clocks inside the driver (i.e. create clk_null
> in clock driver), then we don't need to model the dummy clock in device.
> Is it an acceptable way?
> 
> > So the clock tree should be modeled according to how the hardware is layed
> > out not how you want to use the clocks at the moment :-) .
> > 
> > It would it any case be good, if you could describe where these clocks
> > come
> > from in the hardware, so we can find the best solution on how to model
> > those.
> In fact, I don't know where these clocks come from at all, especially
> when they come from outside of SoC. Besides, some clocks don't need to
> model in CCF, but they can be the source of clocks that controlled by
> CCF.

If a clock is used inside the ccf driver, its tree should be modeled according 
to the hardware - including these external clocks. Somebody (at least some 
chip designer or so) should know where these clocks actually come from ;-) .


> I don't think ALL clocks on a SoC need to be handled in CCF, so there
> should be some clocks don't have a "real" or "correct" parent. In
> current implementation I use a dummy clock (clk_null) to be the unreal
> parent.

You are right that not all clocks needs to be implemented in a ccf _driver_, 
but as the devicetree binding describes the hardware and is supposed to be 
stable and (nearly) unchangeable outside-connections of the clock block need 
to be defined carefully.


> Do you think we should model as more clocks as we can in CCF even we
> don't need them? If no, how do we handle those clocks which are not
> handled in CCF but can be parent of CCF clocks?

In general I think everything that has a connection to the outside (external 
source clock and clocks used by soc ips) should be modeled precisely. What 
then happens inside the clock controller is less important, as it normally can 
be fixed later on too.


Citing my own code [which got inspired on how Samsung did this], Rockchip 
clock controllers also have numerous possible external clock inputs with only 
the 24MHz oscillator being required [0].

All the other clocks may or may not be present. For example xin32k normally 
comes from an i2c-connected rtc or pmic chip which gets probed a lot later in 
the boot process, so we rely on the orphan-handling of the ccf for these.


So please really try to find out what these clocks are in the first place ... 
somebody must know this ;-)


Heiko


[0] 
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/Documentation/devicetree/bindings/clock/rockchip,rk3288-cru.txt#n26

[PATCH] arm64: dts: mt8173: add clock_null

[James Liao]

On Wed, 2015-06-24 at 12:24 +0200, Heiko St?bner wrote:
> Am Mittwoch, 24. Juni 2015, 15:54:15 schrieb James Liao:
> > On Mon, 2015-06-22 at 14:53 +0200, Heiko St?bner wrote:
> > > Am Montag, 22. Juni 2015, 11:38:37 schrieb James Liao:
> > > > On Fri, 2015-06-19 at 13:36 +0200, Heiko Stuebner wrote:
> > > > Some clocks such as clkph_mck_o, we don't really care where they come
> > > > from and what frequencies are. We model these clocks just because they
> > > > or their derived clocks can be the source of topckgen muxes. Is there a
> > > > better way to model "don't care" clocks?
> > > 
> > > There are two different concepts at work here. You might not care in your
> > > kernel driver implementation _at the moment_ where the clocks come from;
> > > but the devicetree is supposed to model how the hardware is structured
> > > and not contain implementation specific details.
> > 
> > If we model "don't care" clocks inside the driver (i.e. create clk_null
> > in clock driver), then we don't need to model the dummy clock in device.
> > Is it an acceptable way?
> > 
> > > So the clock tree should be modeled according to how the hardware is layed
> > > out not how you want to use the clocks at the moment :-) .
> > > 
> > > It would it any case be good, if you could describe where these clocks
> > > come
> > > from in the hardware, so we can find the best solution on how to model
> > > those.
> > In fact, I don't know where these clocks come from at all, especially
> > when they come from outside of SoC. Besides, some clocks don't need to
> > model in CCF, but they can be the source of clocks that controlled by
> > CCF.
> 
> If a clock is used inside the ccf driver, its tree should be modeled according 
> to the hardware - including these external clocks. Somebody (at least some 
> chip designer or so) should know where these clocks actually come from ;-) .
> 
> 
> > I don't think ALL clocks on a SoC need to be handled in CCF, so there
> > should be some clocks don't have a "real" or "correct" parent. In
> > current implementation I use a dummy clock (clk_null) to be the unreal
> > parent.
> 
> You are right that not all clocks needs to be implemented in a ccf _driver_, 
> but as the devicetree binding describes the hardware and is supposed to be 
> stable and (nearly) unchangeable outside-connections of the clock block need 
> to be defined carefully.
> 
> 
> > Do you think we should model as more clocks as we can in CCF even we
> > don't need them? If no, how do we handle those clocks which are not
> > handled in CCF but can be parent of CCF clocks?
> 
> In general I think everything that has a connection to the outside (external 
> source clock and clocks used by soc ips) should be modeled precisely. What 
> then happens inside the clock controller is less important, as it normally can 
> be fixed later on too.
> 
> 
> Citing my own code [which got inspired on how Samsung did this], Rockchip 
> clock controllers also have numerous possible external clock inputs with only 
> the 24MHz oscillator being required [0].
> 
> All the other clocks may or may not be present. For example xin32k normally 
> comes from an i2c-connected rtc or pmic chip which gets probed a lot later in 
> the boot process, so we rely on the orphan-handling of the ccf for these.
> 
> 
> So please really try to find out what these clocks are in the first place ... 
> somebody must know this ;-)
> 
> 
> Heiko
> 
> 
> [0] 
> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/Documentation/devicetree/bindings/clock/rockchip,rk3288-cru.txt#n26

Hi Heiko,

There are 4 clocks which are derived from clk_null directly in current
topckgen implementation:

	clkph_mck_o, dpi_ck, usb_syspll_125m, hdmitx_dig_cts

Our designer mentioned 2 things about external clocks:

1. These 4 clocks come from analog macro, not from PLL, nor from
external clocks directly.

2. MT8173 only has 2 external clocks: 26M and 32K.

Analog macro may refer to 26M or other clocks to generate clocks with
specified frequency. But we don't know and don't care how they generate
these clocks. So we want to use a dummy clock to be the root of these 4
clocks.


Hi Sascha,

Do you have any suggestion on clk_null?


Best regards,

James

[PATCH] arm64: dts: mt8173: add clock_null

[Sascha Hauer]

On Tue, Jun 30, 2015 at 05:07:09PM +0800, James Liao wrote:
> Hi Heiko,
> 
> There are 4 clocks which are derived from clk_null directly in current
> topckgen implementation:
> 
> 	clkph_mck_o, dpi_ck, usb_syspll_125m, hdmitx_dig_cts
> 
> Our designer mentioned 2 things about external clocks:
> 
> 1. These 4 clocks come from analog macro, not from PLL, nor from
> external clocks directly.

Ok, this means there actually are clocks. We can't control these clock and
they have some known or unknown rate. This makes them fixed clocks. Just
specify them in the device tree and you are done. Give them reasonable
names and the rate if you know it, 0 otherwise.

The problem is not that you use fixed clocks for non software
controllable clocks of unknwown rates, but that you try to use a single
clock for all of them and name it 'dummy' or 'null'. Name it

dpi_ck {
	compatible = "fixed-clock";
	rate = <0>; /* unknown, generated by some Analog block */
};

Technically it's the same, but it sounds much more professional and like
you know what you are doing ;)

Sascha

-- 
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

[PATCH] arm64: dts: mt8173: add clock_null

[Daniel Kurtz]

On Wed, Jul 1, 2015 at 2:49 PM, Sascha Hauer <s.hauer@pengutronix.de> wrote:
>
> On Tue, Jun 30, 2015 at 05:07:09PM +0800, James Liao wrote:
> > Hi Heiko,
> >
> > There are 4 clocks which are derived from clk_null directly in current
> > topckgen implementation:
> >
> >       clkph_mck_o, dpi_ck, usb_syspll_125m, hdmitx_dig_cts
> >
> > Our designer mentioned 2 things about external clocks:
> >
> > 1. These 4 clocks come from analog macro, not from PLL, nor from
> > external clocks directly.
>
> Ok, this means there actually are clocks. We can't control these clock and
> they have some known or unknown rate. This makes them fixed clocks. Just
> specify them in the device tree and you are done. Give them reasonable
> names and the rate if you know it, 0 otherwise.
>
> The problem is not that you use fixed clocks for non software
> controllable clocks of unknwown rates, but that you try to use a single
> clock for all of them and name it 'dummy' or 'null'. Name it
>
> dpi_ck {
>         compatible = "fixed-clock";
>         rate = <0>; /* unknown, generated by some Analog block */
> };

It would be nice, though, to use the real clock rates.
Otherwise, we end up with a bunch of unknown clock rates, like this:

   clock                         enable_cnt  prepare_cnt        rate
accuracy   phase
----------------------------------------------------------------------------------------
 clk_null                                 2            2           0
       0 0
    mm_lvds_cts                           0            0           0
       0 0
    mm_lvds_pixel                         0            0           0
       0 0
    mm_dpi1_pixel                         0            0           0
       0 0
    mm_dsi1_digital                       0            0           0
       0 0
    mm_dsi0_digital                       1            1           0
       0 0
    infra_cpum                            0            0           0
       0 0
    hdmitx_dig_cts                        0            0           0
       0 0
       hdmi_sel                           0            0           0
       0 0
          mm_hdmi_pllck                   0            0           0
       0 0
       hdmitxpll_d3                       0            0           0
       0 0
       hdmitxpll_d2                       0            0           0
       0 0
    usb_syspll_125m                       0            0           0
       0 0
    dpi_ck                                0            0           0
       0 0
    clkph_mck_o                           1            1           0
       0 0
       dmpll_d16                          0            0           0
       0 0
       dmpll_d8                           0            0           0
       0 0
       dmpll_d4                           0            0           0
       0 0
       dmpll_d2                           0            0           0
       0 0
       dmpll_ck                           1            1           0
       0 0
          mem_sel                         2            2           0
       0 0
             infra_m4u                    1            1           0
       0 0

Furthermore, at least some of these children clocks are possible
source clocks for other clocks for which we do want to know the
resulting frequency.  For example, the "dmpll_*" clocks are mux inputs
for many of the subsystem clocks.

-Dan

>
> Technically it's the same, but it sounds much more professional and like
> you know what you are doing ;)
>
> Sascha
>
> --
> Pengutronix e.K.                           |                             |
> Industrial Linux Solutions                 | http://www.pengutronix.de/  |
> Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
> Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo at vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

[PATCH] arm64: dts: mt8173: add clock_null

[James Liao]

Hi Daniel,

On Wed, 2015-07-01 at 19:54 +0800, Daniel Kurtz wrote:
> On Wed, Jul 1, 2015 at 2:49 PM, Sascha Hauer <s.hauer@pengutronix.de> wrote:
> > The problem is not that you use fixed clocks for non software
> > controllable clocks of unknwown rates, but that you try to use a single
> > clock for all of them and name it 'dummy' or 'null'. Name it
> >
> > dpi_ck {
> >         compatible = "fixed-clock";
> >         rate = <0>; /* unknown, generated by some Analog block */
> > };
> 
> It would be nice, though, to use the real clock rates.
> Otherwise, we end up with a bunch of unknown clock rates, like this:
> 
>    clock                         enable_cnt  prepare_cnt        rate
> accuracy   phase
> ----------------------------------------------------------------------------------------
>  clk_null                                 2            2           0
>        0 0
>     mm_lvds_cts                           0            0           0
>        0 0
>     mm_lvds_pixel                         0            0           0
>        0 0
>     mm_dpi1_pixel                         0            0           0
>        0 0

> Furthermore, at least some of these children clocks are possible
> source clocks for other clocks for which we do want to know the
> resulting frequency.  For example, the "dmpll_*" clocks are mux inputs
> for many of the subsystem clocks.

These clocks such as clkph_mck_o are configured by other modules before
kernel init, and their rates may different among platforms. So we can't
use a hard-coded rate for them. And we also don't care the actual rate
of these clocks, so assign a dummy rate such as 0 to them should be a
better way.


Best regards,

james

[PATCH] arm64: dts: mt8173: add clock_null

[Daniel Kurtz]

On Thu, Jul 2, 2015 at 11:06 AM, James Liao <jamesjj.liao@mediatek.com> wrote:
> Hi Daniel,
>
> On Wed, 2015-07-01 at 19:54 +0800, Daniel Kurtz wrote:
>> On Wed, Jul 1, 2015 at 2:49 PM, Sascha Hauer <s.hauer@pengutronix.de> wrote:
>> > The problem is not that you use fixed clocks for non software
>> > controllable clocks of unknwown rates, but that you try to use a single
>> > clock for all of them and name it 'dummy' or 'null'. Name it
>> >
>> > dpi_ck {
>> >         compatible = "fixed-clock";
>> >         rate = <0>; /* unknown, generated by some Analog block */
>> > };
>>
>> It would be nice, though, to use the real clock rates.
>> Otherwise, we end up with a bunch of unknown clock rates, like this:
>>
>>    clock                         enable_cnt  prepare_cnt        rate
>> accuracy   phase
>> ----------------------------------------------------------------------------------------
>>  clk_null                                 2            2           0
>>        0 0
>>     mm_lvds_cts                           0            0           0
>>        0 0
>>     mm_lvds_pixel                         0            0           0
>>        0 0
>>     mm_dpi1_pixel                         0            0           0
>>        0 0
>
>> Furthermore, at least some of these children clocks are possible
>> source clocks for other clocks for which we do want to know the
>> resulting frequency.  For example, the "dmpll_*" clocks are mux inputs
>> for many of the subsystem clocks.
>
> These clocks such as clkph_mck_o are configured by other modules before
> kernel init, and their rates may different among platforms.

What other modules?
Do you mean the rates are configured by firmware?
How are the rates set?
Are there registers that configure its rate?
If so, why can't the kernel read these registers and compute the current rate?


For mt8173, we are essentially discussing the following clocks (whose
sole parent is clk_null):

FACTOR(CLK_TOP_CLKPH_MCK_O, "clkph_mck_o", "clk_null", 1, 1),
FACTOR(CLK_TOP_DPI, "dpi_ck", "clk_null", 1, 1),
FACTOR(CLK_TOP_USB_SYSPLL_125M, "usb_syspll_125m", "clk_null", 1, 1),
FACTOR(CLK_TOP_HDMITX_DIG_CTS, "hdmitx_dig_cts", "clk_null", 1, 1),
GATE_ICG(CLK_INFRA_CPUM, "infra_cpum", "clk_null", 15),
GATE_MM1(CLK_MM_DSI0_DIGITAL, "mm_dsi0_digital", "clk_null", 5),
GATE_MM1(CLK_MM_DSI1_DIGITAL, "mm_dsi1_digital", "clk_null", 7),
GATE_MM1(CLK_MM_DPI1_PIXEL, "mm_dpi1_pixel", "clk_null", 10),
GATE_MM1(CLK_MM_LVDS_PIXEL, "mm_lvds_pixel", "clk_null", 16),
GATE_MM1(CLK_MM_LVDS_CTS, "mm_lvds_cts", "clk_null", 17),

clkph_mck_o - This is the parent for dmpll_*, which are themselves
(potential) parent clocks for nearly every subsystem.
In fact, as shown above, the dmpll_* is the selected parent for
several other clocks, which all end up with an unknown rate.
So, I think it is worth investigating a little more how to properly
read or otherwise specify the rate for clkph_mck_o.

dpi_ck, infra_cpum, mm_dsi0_digital, mm_dsi1_digital, mm_lvds_cts -
These are a dead-end (internal?) clock.
It is probably fine if their rates are unknown (0 Hz).

usb_syspll_125m - This sounds like a fixed 125 MHz clock.  It is also
a possible parent usb30 clock, so its value will propagate.

hdmitx_dig_cts - This is the root clock for the tree leading to
mm_hdmi_pllck, which includes hdmitxpll_d* and hdmi_sel.
However, I don't know how "mm_hdmi_pllck" is used.

mm_dpi1_pixel, mm_lvds_pixel - These two look very suspicious.  The
similar "mm_dpi0_pixel" and "mm_hdmi_pixel" have parent dpi0_sel.
It looks like maybe they should have "dpi1_sel" or "dpilvds_sel" as
parents, but the _sel are not hooked up.

-Dan

> So we can't
> use a hard-coded rate for them. And we also don't care the actual rate
> of these clocks, so assign a dummy rate such as 0 to them should be a
> better way.
>
>
> Best regards,
>
> james

[PATCH] arm64: dts: mt8173: add clock_null

[James Liao]

On Thu, 2015-07-02 at 12:23 +0800, Daniel Kurtz wrote:
> On Thu, Jul 2, 2015 at 11:06 AM, James Liao <jamesjj.liao@mediatek.com> wrote:
> > These clocks such as clkph_mck_o are configured by other modules before
> > kernel init, and their rates may different among platforms.
> 
> What other modules?
> Do you mean the rates are configured by firmware?
> How are the rates set?
> Are there registers that configure its rate?
> If so, why can't the kernel read these registers and compute the current rate?

CLKPH_MCK_O for example, it's a DRAM related clock, and initialized in
preloader (before kernel init). It's setting may be different because
using different DRAM module. And the setting may be changed dynamically
in runtime.

We don't care CLKPH_MCK_O's rate in CCF because we don't need to change
mem_sel's setting in kernel, and there is not driver need to know
mem_sel's rate.

> 
> For mt8173, we are essentially discussing the following clocks (whose
> sole parent is clk_null):
> 
> FACTOR(CLK_TOP_CLKPH_MCK_O, "clkph_mck_o", "clk_null", 1, 1),
> FACTOR(CLK_TOP_DPI, "dpi_ck", "clk_null", 1, 1),
> FACTOR(CLK_TOP_USB_SYSPLL_125M, "usb_syspll_125m", "clk_null", 1, 1),
> FACTOR(CLK_TOP_HDMITX_DIG_CTS, "hdmitx_dig_cts", "clk_null", 1, 1),
> GATE_ICG(CLK_INFRA_CPUM, "infra_cpum", "clk_null", 15),
> GATE_MM1(CLK_MM_DSI0_DIGITAL, "mm_dsi0_digital", "clk_null", 5),
> GATE_MM1(CLK_MM_DSI1_DIGITAL, "mm_dsi1_digital", "clk_null", 7),
> GATE_MM1(CLK_MM_DPI1_PIXEL, "mm_dpi1_pixel", "clk_null", 10),
> GATE_MM1(CLK_MM_LVDS_PIXEL, "mm_lvds_pixel", "clk_null", 16),
> GATE_MM1(CLK_MM_LVDS_CTS, "mm_lvds_cts", "clk_null", 17),
> 
> clkph_mck_o - This is the parent for dmpll_*, which are themselves
> (potential) parent clocks for nearly every subsystem.
> In fact, as shown above, the dmpll_* is the selected parent for
> several other clocks, which all end up with an unknown rate.
> So, I think it is worth investigating a little more how to properly
> read or otherwise specify the rate for clkph_mck_o.

Please see above.

> dpi_ck, infra_cpum, mm_dsi0_digital, mm_dsi1_digital, mm_lvds_cts -
> These are a dead-end (internal?) clock.
> It is probably fine if their rates are unknown (0 Hz).
> 
> usb_syspll_125m - This sounds like a fixed 125 MHz clock.  It is also
> a possible parent usb30 clock, so its value will propagate.
> 
> hdmitx_dig_cts - This is the root clock for the tree leading to
> mm_hdmi_pllck, which includes hdmitxpll_d* and hdmi_sel.
> However, I don't know how "mm_hdmi_pllck" is used.
> 
> mm_dpi1_pixel, mm_lvds_pixel - These two look very suspicious.  The
> similar "mm_dpi0_pixel" and "mm_hdmi_pixel" have parent dpi0_sel.
> It looks like maybe they should have "dpi1_sel" or "dpilvds_sel" as
> parents, but the _sel are not hooked up.

Subsystem clocks with parent clk_null may have different reasons. I'll
get back to you later.


Best regards,

James

[PATCH] arm64: dts: mt8173: add clock_null

[Heiko Stübner]

Am Freitag, 3. Juli 2015, 15:45:12 schrieb James Liao:
> On Thu, 2015-07-02 at 12:23 +0800, Daniel Kurtz wrote:
> > On Thu, Jul 2, 2015 at 11:06 AM, James Liao <jamesjj.liao@mediatek.com> 
wrote:
> > > These clocks such as clkph_mck_o are configured by other modules before
> > > kernel init, and their rates may different among platforms.
> > 
> > What other modules?
> > Do you mean the rates are configured by firmware?
> > How are the rates set?
> > Are there registers that configure its rate?
> > If so, why can't the kernel read these registers and compute the current
> > rate?
> CLKPH_MCK_O for example, it's a DRAM related clock, and initialized in
> preloader (before kernel init). It's setting may be different because
> using different DRAM module. And the setting may be changed dynamically
> in runtime.

If it's set in the preloader (and maybe changed at runtime) this means, its 
setting can also be read back to determine its current clock rate, so this is 
a non-argument here ;-)


> We don't care CLKPH_MCK_O's rate in CCF because we don't need to change
> mem_sel's setting in kernel, and there is not driver need to know
> mem_sel's rate.

As Daniel said, some of these clocks may be parents of other clocks, and not 
knowing their rate might hurt in the future - especially when it's easy after 
all to get its actual value like in your CLKPH_MCK_O example above.


Heiko


> > For mt8173, we are essentially discussing the following clocks (whose
> > sole parent is clk_null):
> > 
> > FACTOR(CLK_TOP_CLKPH_MCK_O, "clkph_mck_o", "clk_null", 1, 1),
> > FACTOR(CLK_TOP_DPI, "dpi_ck", "clk_null", 1, 1),
> > FACTOR(CLK_TOP_USB_SYSPLL_125M, "usb_syspll_125m", "clk_null", 1, 1),
> > FACTOR(CLK_TOP_HDMITX_DIG_CTS, "hdmitx_dig_cts", "clk_null", 1, 1),
> > GATE_ICG(CLK_INFRA_CPUM, "infra_cpum", "clk_null", 15),
> > GATE_MM1(CLK_MM_DSI0_DIGITAL, "mm_dsi0_digital", "clk_null", 5),
> > GATE_MM1(CLK_MM_DSI1_DIGITAL, "mm_dsi1_digital", "clk_null", 7),
> > GATE_MM1(CLK_MM_DPI1_PIXEL, "mm_dpi1_pixel", "clk_null", 10),
> > GATE_MM1(CLK_MM_LVDS_PIXEL, "mm_lvds_pixel", "clk_null", 16),
> > GATE_MM1(CLK_MM_LVDS_CTS, "mm_lvds_cts", "clk_null", 17),
> > 
> > clkph_mck_o - This is the parent for dmpll_*, which are themselves
> > (potential) parent clocks for nearly every subsystem.
> > In fact, as shown above, the dmpll_* is the selected parent for
> > several other clocks, which all end up with an unknown rate.
> > So, I think it is worth investigating a little more how to properly
> > read or otherwise specify the rate for clkph_mck_o.
> 
> Please see above.
> 
> > dpi_ck, infra_cpum, mm_dsi0_digital, mm_dsi1_digital, mm_lvds_cts -
> > These are a dead-end (internal?) clock.
> > It is probably fine if their rates are unknown (0 Hz).
> > 
> > usb_syspll_125m - This sounds like a fixed 125 MHz clock.  It is also
> > a possible parent usb30 clock, so its value will propagate.
> > 
> > hdmitx_dig_cts - This is the root clock for the tree leading to
> > mm_hdmi_pllck, which includes hdmitxpll_d* and hdmi_sel.
> > However, I don't know how "mm_hdmi_pllck" is used.
> > 
> > mm_dpi1_pixel, mm_lvds_pixel - These two look very suspicious.  The
> > similar "mm_dpi0_pixel" and "mm_hdmi_pixel" have parent dpi0_sel.
> > It looks like maybe they should have "dpi1_sel" or "dpilvds_sel" as
> > parents, but the _sel are not hooked up.
> 
> Subsystem clocks with parent clk_null may have different reasons. I'll
> get back to you later.
> 
> 
> Best regards,
> 
> James

[PATCH] arm64: dts: mt8173: add clock_null

[James Liao]

Hi Sascha,

On Wed, 2015-07-01 at 08:49 +0200, Sascha Hauer wrote:
> On Tue, Jun 30, 2015 at 05:07:09PM +0800, James Liao wrote:
> > There are 4 clocks which are derived from clk_null directly in current
> > topckgen implementation:
> > 
> > 	clkph_mck_o, dpi_ck, usb_syspll_125m, hdmitx_dig_cts
> > 
> > Our designer mentioned 2 things about external clocks:
> > 
> > 1. These 4 clocks come from analog macro, not from PLL, nor from
> > external clocks directly.
> 
> Ok, this means there actually are clocks. We can't control these clock and
> they have some known or unknown rate. This makes them fixed clocks. Just
> specify them in the device tree and you are done. Give them reasonable
> names and the rate if you know it, 0 otherwise.
> 
> The problem is not that you use fixed clocks for non software
> controllable clocks of unknwown rates, but that you try to use a single
> clock for all of them and name it 'dummy' or 'null'. Name it
> 
> dpi_ck {
> 	compatible = "fixed-clock";
> 	rate = <0>; /* unknown, generated by some Analog block */
> };
> 
> Technically it's the same, but it sounds much more professional and like
> you know what you are doing ;)

These clocks are SoC internal clocks. Is it suitable to specify them in
the device tree?

According to your and Heiko's suggestion, it looks the best way should
be specifying these clocks in the driver code without a dummy parent.
i.e.:

Original code:

FACTOR(CLK_TOP_CLKPH_MCK_O, "clkph_mck_o", "clk_null", 1, 1),
FACTOR(CLK_TOP_DPI, "dpi_ck", "clk_null", 1, 1),
FACTOR(CLK_TOP_USB_SYSPLL_125M, "usb_syspll_125m", "clk_null", 1, 1),
FACTOR(CLK_TOP_HDMITX_DIG_CTS, "hdmitx_dig_cts", "clk_null", 1, 1),

New code:

FIXED(CLK_TOP_CLKPH_MCK_O, "clkph_mck_o", UNKNOWN_RATE),
FIXED(CLK_TOP_DPI, "dpi_ck", UNKNOWN_RATE),
FIXED(CLK_TOP_USB_SYSPLL_125M, "usb_syspll_125m", UNKNOWN_RATE),
FIXED(CLK_TOP_HDMITX_DIG_CTS, "hdmitx_dig_cts", UNKNOWN_RATE),

Then we don't need to specify a dummy clock such as clk_null in device
tree.


Best regards,

James

[PATCH] arm64: dts: mt8173: add clock_null

[Sascha Hauer]

On Thu, Jun 18, 2015 at 01:29:11PM +0800, Eddie Huang wrote:
> Add clk_null, which represents clocks that can not / need not
> controlled by software.
> There are many clocks' parent set to clk_null.
> 
> Signed-off-by: James Liao <jamesjj.liao@mediatek.com>
> Signed-off-by: Eddie Huang <eddie.huang@mediatek.com>
> ---
> Base on 4.1-rc1
> 
> Change-Id: I4db9b40d07e28f54f7bae9b676316cbd6a962124
> ---
>  arch/arm64/boot/dts/mediatek/mt8173.dtsi | 6 ++++++
>  1 file changed, 6 insertions(+)
> 
> diff --git a/arch/arm64/boot/dts/mediatek/mt8173.dtsi b/arch/arm64/boot/dts/mediatek/mt8173.dtsi
> index 924fdb6..4798f44 100644
> --- a/arch/arm64/boot/dts/mediatek/mt8173.dtsi
> +++ b/arch/arm64/boot/dts/mediatek/mt8173.dtsi
> @@ -81,6 +81,12 @@
>  		cpu_on	      = <0x84000003>;
>  	};
>  
> +	clk_null: clk_null {
> +		compatible = "fixed-clock";
> +		clock-frequency = <0>;
> +		#clock-cells = <0>;
> +	};

The discussion around this patch shows that we don't want to have this
clock in the device tree as it is not a hardware description.

Ok, fine. Eddie, you told us that the rate of the current clk_null children
is not interesting. What's the motivation to send this patch anyway
then? Why can't you keep its children on the orphan list where they are
already now?

Another possibility would be to instantiate the clk_null clock from C
code rather than from the device tree. This way we wouldn't put any
wrong descriptions into the device tree and still can implement the
support for the real parent clocks when we actually need them.

Sascha

-- 
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

[PATCH] arm64: dts: mt8173: add clock_null

[Daniel Kurtz]

On Tue, Jul 7, 2015 at 9:07 PM, Sascha Hauer <s.hauer@pengutronix.de> wrote:
> On Thu, Jun 18, 2015 at 01:29:11PM +0800, Eddie Huang wrote:
>> Add clk_null, which represents clocks that can not / need not
>> controlled by software.
>> There are many clocks' parent set to clk_null.
>>
>> Signed-off-by: James Liao <jamesjj.liao@mediatek.com>
>> Signed-off-by: Eddie Huang <eddie.huang@mediatek.com>
>> ---
>> Base on 4.1-rc1
>>
>> Change-Id: I4db9b40d07e28f54f7bae9b676316cbd6a962124
>> ---
>>  arch/arm64/boot/dts/mediatek/mt8173.dtsi | 6 ++++++
>>  1 file changed, 6 insertions(+)
>>
>> diff --git a/arch/arm64/boot/dts/mediatek/mt8173.dtsi b/arch/arm64/boot/dts/mediatek/mt8173.dtsi
>> index 924fdb6..4798f44 100644
>> --- a/arch/arm64/boot/dts/mediatek/mt8173.dtsi
>> +++ b/arch/arm64/boot/dts/mediatek/mt8173.dtsi
>> @@ -81,6 +81,12 @@
>>               cpu_on        = <0x84000003>;
>>       };
>>
>> +     clk_null: clk_null {
>> +             compatible = "fixed-clock";
>> +             clock-frequency = <0>;
>> +             #clock-cells = <0>;
>> +     };
>
> The discussion around this patch shows that we don't want to have this
> clock in the device tree as it is not a hardware description.
>
> Ok, fine. Eddie, you told us that the rate of the current clk_null children
> is not interesting. What's the motivation to send this patch anyway
> then? Why can't you keep its children on the orphan list where they are
> already now?
>
> Another possibility would be to instantiate the clk_null clock from C
> code rather than from the device tree. This way we wouldn't put any
> wrong descriptions into the device tree and still can implement the
> support for the real parent clocks when we actually need them.

Some device nodes, like mmc, use a clk_null phandle as one of their clocks:

mmc1: mmc at 11240000 {
        compatible = "mediatek,mt8173-mmc",
                     "mediatek,mt8135-mmc";
        reg = <0 0x11240000 0 0x1000>;
        interrupts = <GIC_SPI 72 IRQ_TYPE_LEVEL_LOW>;
        clocks = <&pericfg CLK_PERI_MSDC30_1>,
                 <&clk_null>;
        clock-names = "source", "hclk";
        status = "disabled";
};

-Dan


> Sascha
>
> --
> Pengutronix e.K.                           |                             |
> Industrial Linux Solutions                 | http://www.pengutronix.de/  |
> Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
> Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo at vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

[PATCH] arm64: dts: mt8173: add clock_null

[Sascha Hauer]

On Tue, Jul 07, 2015 at 10:15:29PM +0800, Daniel Kurtz wrote:
> On Tue, Jul 7, 2015 at 9:07 PM, Sascha Hauer <s.hauer@pengutronix.de> wrote:
> > On Thu, Jun 18, 2015 at 01:29:11PM +0800, Eddie Huang wrote:
> >> Add clk_null, which represents clocks that can not / need not
> >> controlled by software.
> >> There are many clocks' parent set to clk_null.
> >>
> >> Signed-off-by: James Liao <jamesjj.liao@mediatek.com>
> >> Signed-off-by: Eddie Huang <eddie.huang@mediatek.com>
> >> ---
> >> Base on 4.1-rc1
> >>
> >> Change-Id: I4db9b40d07e28f54f7bae9b676316cbd6a962124
> >> ---
> >>  arch/arm64/boot/dts/mediatek/mt8173.dtsi | 6 ++++++
> >>  1 file changed, 6 insertions(+)
> >>
> >> diff --git a/arch/arm64/boot/dts/mediatek/mt8173.dtsi b/arch/arm64/boot/dts/mediatek/mt8173.dtsi
> >> index 924fdb6..4798f44 100644
> >> --- a/arch/arm64/boot/dts/mediatek/mt8173.dtsi
> >> +++ b/arch/arm64/boot/dts/mediatek/mt8173.dtsi
> >> @@ -81,6 +81,12 @@
> >>               cpu_on        = <0x84000003>;
> >>       };
> >>
> >> +     clk_null: clk_null {
> >> +             compatible = "fixed-clock";
> >> +             clock-frequency = <0>;
> >> +             #clock-cells = <0>;
> >> +     };
> >
> > The discussion around this patch shows that we don't want to have this
> > clock in the device tree as it is not a hardware description.
> >
> > Ok, fine. Eddie, you told us that the rate of the current clk_null children
> > is not interesting. What's the motivation to send this patch anyway
> > then? Why can't you keep its children on the orphan list where they are
> > already now?
> >
> > Another possibility would be to instantiate the clk_null clock from C
> > code rather than from the device tree. This way we wouldn't put any
> > wrong descriptions into the device tree and still can implement the
> > support for the real parent clocks when we actually need them.
> 
> Some device nodes, like mmc, use a clk_null phandle as one of their clocks:
> 
> mmc1: mmc at 11240000 {
>         compatible = "mediatek,mt8173-mmc",
>                      "mediatek,mt8135-mmc";
>         reg = <0 0x11240000 0 0x1000>;
>         interrupts = <GIC_SPI 72 IRQ_TYPE_LEVEL_LOW>;
>         clocks = <&pericfg CLK_PERI_MSDC30_1>,
>                  <&clk_null>;
>         clock-names = "source", "hclk";
>         status = "disabled";
> };

This is another case than the one we discussed about. In the case above
I motivated using a dummy clock since the clock exists in the system,
but is not software controllable. To abstract this from the driver
(which needs this clock since it exists) we here have the dummy clock.
However, of course I can't prove the clock is indeed not software
controllable; that's only the information I have.

Sascha

-- 
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

[PATCH] arm64: dts: mt8173: add clock_null

[Daniel Kurtz]

On Tue, Jul 7, 2015 at 10:36 PM, Sascha Hauer <s.hauer@pengutronix.de> wrote:
> On Tue, Jul 07, 2015 at 10:15:29PM +0800, Daniel Kurtz wrote:
>> On Tue, Jul 7, 2015 at 9:07 PM, Sascha Hauer <s.hauer@pengutronix.de> wrote:
>> > On Thu, Jun 18, 2015 at 01:29:11PM +0800, Eddie Huang wrote:
>> >> Add clk_null, which represents clocks that can not / need not
>> >> controlled by software.
>> >> There are many clocks' parent set to clk_null.
>> >>
>> >> Signed-off-by: James Liao <jamesjj.liao@mediatek.com>
>> >> Signed-off-by: Eddie Huang <eddie.huang@mediatek.com>
>> >> ---
>> >> Base on 4.1-rc1
>> >>
>> >> Change-Id: I4db9b40d07e28f54f7bae9b676316cbd6a962124
>> >> ---
>> >>  arch/arm64/boot/dts/mediatek/mt8173.dtsi | 6 ++++++
>> >>  1 file changed, 6 insertions(+)
>> >>
>> >> diff --git a/arch/arm64/boot/dts/mediatek/mt8173.dtsi b/arch/arm64/boot/dts/mediatek/mt8173.dtsi
>> >> index 924fdb6..4798f44 100644
>> >> --- a/arch/arm64/boot/dts/mediatek/mt8173.dtsi
>> >> +++ b/arch/arm64/boot/dts/mediatek/mt8173.dtsi
>> >> @@ -81,6 +81,12 @@
>> >>               cpu_on        = <0x84000003>;
>> >>       };
>> >>
>> >> +     clk_null: clk_null {
>> >> +             compatible = "fixed-clock";
>> >> +             clock-frequency = <0>;
>> >> +             #clock-cells = <0>;
>> >> +     };
>> >
>> > The discussion around this patch shows that we don't want to have this
>> > clock in the device tree as it is not a hardware description.
>> >
>> > Ok, fine. Eddie, you told us that the rate of the current clk_null children
>> > is not interesting. What's the motivation to send this patch anyway
>> > then? Why can't you keep its children on the orphan list where they are
>> > already now?
>> >
>> > Another possibility would be to instantiate the clk_null clock from C
>> > code rather than from the device tree. This way we wouldn't put any
>> > wrong descriptions into the device tree and still can implement the
>> > support for the real parent clocks when we actually need them.
>>
>> Some device nodes, like mmc, use a clk_null phandle as one of their clocks:
>>
>> mmc1: mmc at 11240000 {
>>         compatible = "mediatek,mt8173-mmc",
>>                      "mediatek,mt8135-mmc";
>>         reg = <0 0x11240000 0 0x1000>;
>>         interrupts = <GIC_SPI 72 IRQ_TYPE_LEVEL_LOW>;
>>         clocks = <&pericfg CLK_PERI_MSDC30_1>,
>>                  <&clk_null>;
>>         clock-names = "source", "hclk";
>>         status = "disabled";
>> };
>
> This is another case than the one we discussed about. In the case above
> I motivated using a dummy clock since the clock exists in the system,
> but is not software controllable. To abstract this from the driver
> (which needs this clock since it exists) we here have the dummy clock.
> However, of course I can't prove the clock is indeed not software
> controllable; that's only the information I have.

I was trying to answer your question "What's the motivation to send
this patch anyway?".
The motivation is to send follow on patches that use the clk_null
phandle.  We need to provide some clock as the mmc1's hclk.  I do not
understand why this has to be "clk_null", though.  It seems like this
should be a real clock coming from one of the real clock_controller
nodes.  After all, the mmc driver is going to be enabling/disabling
this clock for power savings at runtime.  What does that even mean for
clk_null ?

Sorry, I'm not exactly sure what you are saying in your last reply.

>
> Sascha
>
> --
> Pengutronix e.K.                           |                             |
> Industrial Linux Solutions                 | http://www.pengutronix.de/  |
> Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
> Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

[PATCH] arm64: dts: mt8173: add clock_null

[Eddie Huang]

On Tue, 2015-07-07 at 23:10 +0800, Daniel Kurtz wrote:
> On Tue, Jul 7, 2015 at 10:36 PM, Sascha Hauer <s.hauer@pengutronix.de> wrote:
> > On Tue, Jul 07, 2015 at 10:15:29PM +0800, Daniel Kurtz wrote:
> >> On Tue, Jul 7, 2015 at 9:07 PM, Sascha Hauer <s.hauer@pengutronix.de> wrote:
> >> > On Thu, Jun 18, 2015 at 01:29:11PM +0800, Eddie Huang wrote:
> >> >> Add clk_null, which represents clocks that can not / need not
> >> >> controlled by software.
> >> >> There are many clocks' parent set to clk_null.
> >> >>
> >> >> Signed-off-by: James Liao <jamesjj.liao@mediatek.com>
> >> >> Signed-off-by: Eddie Huang <eddie.huang@mediatek.com>
> >> >> ---
> >> >> Base on 4.1-rc1
> >> >>
> >> >> Change-Id: I4db9b40d07e28f54f7bae9b676316cbd6a962124
> >> >> ---
> >> >>  arch/arm64/boot/dts/mediatek/mt8173.dtsi | 6 ++++++
> >> >>  1 file changed, 6 insertions(+)
> >> >>
> >> >> diff --git a/arch/arm64/boot/dts/mediatek/mt8173.dtsi b/arch/arm64/boot/dts/mediatek/mt8173.dtsi
> >> >> index 924fdb6..4798f44 100644
> >> >> --- a/arch/arm64/boot/dts/mediatek/mt8173.dtsi
> >> >> +++ b/arch/arm64/boot/dts/mediatek/mt8173.dtsi
> >> >> @@ -81,6 +81,12 @@
> >> >>               cpu_on        = <0x84000003>;
> >> >>       };
> >> >>
> >> >> +     clk_null: clk_null {
> >> >> +             compatible = "fixed-clock";
> >> >> +             clock-frequency = <0>;
> >> >> +             #clock-cells = <0>;
> >> >> +     };
> >> >
> >> > The discussion around this patch shows that we don't want to have this
> >> > clock in the device tree as it is not a hardware description.
> >> >
> >> > Ok, fine. Eddie, you told us that the rate of the current clk_null children
> >> > is not interesting. What's the motivation to send this patch anyway
> >> > then? Why can't you keep its children on the orphan list where they are
> >> > already now?
> >> >
> >> > Another possibility would be to instantiate the clk_null clock from C
> >> > code rather than from the device tree. This way we wouldn't put any
> >> > wrong descriptions into the device tree and still can implement the
> >> > support for the real parent clocks when we actually need them.
> >>
> >> Some device nodes, like mmc, use a clk_null phandle as one of their clocks:
> >>
> >> mmc1: mmc at 11240000 {
> >>         compatible = "mediatek,mt8173-mmc",
> >>                      "mediatek,mt8135-mmc";
> >>         reg = <0 0x11240000 0 0x1000>;
> >>         interrupts = <GIC_SPI 72 IRQ_TYPE_LEVEL_LOW>;
> >>         clocks = <&pericfg CLK_PERI_MSDC30_1>,
> >>                  <&clk_null>;
> >>         clock-names = "source", "hclk";
> >>         status = "disabled";
> >> };
> >
> > This is another case than the one we discussed about. In the case above
> > I motivated using a dummy clock since the clock exists in the system,
> > but is not software controllable. To abstract this from the driver
> > (which needs this clock since it exists) we here have the dummy clock.
> > However, of course I can't prove the clock is indeed not software
> > controllable; that's only the information I have.
> 
> I was trying to answer your question "What's the motivation to send
> this patch anyway?".
> The motivation is to send follow on patches that use the clk_null
> phandle.  We need to provide some clock as the mmc1's hclk.  I do not
> understand why this has to be "clk_null", though.  It seems like this
> should be a real clock coming from one of the real clock_controller
> nodes.  After all, the mmc driver is going to be enabling/disabling
> this clock for power savings at runtime.  What does that even mean for
> clk_null ?

The original purpose of this patch is to provide a common dummy clock
for both software don't care clock and clock that is not software
controllable.I got comments that device tree should describe hardware
and should put exact clock in device tree. I think this is true. So we
will remove this clock_null patch, and:

1. For Mediatek SoC CCF driver, James will clarify clock usage further.
Actually, we still think it's not necessary to describe whole tree that
software don't care, James will deal this in clock driver.

2. For other module that use SW not controllable clock (mmc case
mentioned by Dan), because this is a real clock, we will put a dummy
clock in device tree, like

clk_mmchclk: dummyhclk {
	compatible = "fixed-clock";
	clock-frequency = <0>;
	#clock-cells = <0>;
};

How about this modification ?

Thanks
Eddie

[PATCH] arm64: dts: mt8173: add clock_null

[Sascha Hauer]

On Wed, Jul 08, 2015 at 10:37:21AM +0800, Eddie Huang wrote:
> On Tue, 2015-07-07 at 23:10 +0800, Daniel Kurtz wrote:
> > On Tue, Jul 7, 2015 at 10:36 PM, Sascha Hauer <s.hauer@pengutronix.de> wrote:
> > > On Tue, Jul 07, 2015 at 10:15:29PM +0800, Daniel Kurtz wrote:
> > >> On Tue, Jul 7, 2015 at 9:07 PM, Sascha Hauer <s.hauer@pengutronix.de> wrote:
> > >> > On Thu, Jun 18, 2015 at 01:29:11PM +0800, Eddie Huang wrote:
> > >> >> Add clk_null, which represents clocks that can not / need not
> > >> >> controlled by software.
> > >> >> There are many clocks' parent set to clk_null.
> > >> >>
> > >> >> Signed-off-by: James Liao <jamesjj.liao@mediatek.com>
> > >> >> Signed-off-by: Eddie Huang <eddie.huang@mediatek.com>
> > >> >> ---
> > >> >> Base on 4.1-rc1
> > >> >>
> > >> >> Change-Id: I4db9b40d07e28f54f7bae9b676316cbd6a962124
> > >> >> ---
> > >> >>  arch/arm64/boot/dts/mediatek/mt8173.dtsi | 6 ++++++
> > >> >>  1 file changed, 6 insertions(+)
> > >> >>
> > >> >> diff --git a/arch/arm64/boot/dts/mediatek/mt8173.dtsi b/arch/arm64/boot/dts/mediatek/mt8173.dtsi
> > >> >> index 924fdb6..4798f44 100644
> > >> >> --- a/arch/arm64/boot/dts/mediatek/mt8173.dtsi
> > >> >> +++ b/arch/arm64/boot/dts/mediatek/mt8173.dtsi
> > >> >> @@ -81,6 +81,12 @@
> > >> >>               cpu_on        = <0x84000003>;
> > >> >>       };
> > >> >>
> > >> >> +     clk_null: clk_null {
> > >> >> +             compatible = "fixed-clock";
> > >> >> +             clock-frequency = <0>;
> > >> >> +             #clock-cells = <0>;
> > >> >> +     };
> > >> >
> > >> > The discussion around this patch shows that we don't want to have this
> > >> > clock in the device tree as it is not a hardware description.
> > >> >
> > >> > Ok, fine. Eddie, you told us that the rate of the current clk_null children
> > >> > is not interesting. What's the motivation to send this patch anyway
> > >> > then? Why can't you keep its children on the orphan list where they are
> > >> > already now?
> > >> >
> > >> > Another possibility would be to instantiate the clk_null clock from C
> > >> > code rather than from the device tree. This way we wouldn't put any
> > >> > wrong descriptions into the device tree and still can implement the
> > >> > support for the real parent clocks when we actually need them.
> > >>
> > >> Some device nodes, like mmc, use a clk_null phandle as one of their clocks:
> > >>
> > >> mmc1: mmc at 11240000 {
> > >>         compatible = "mediatek,mt8173-mmc",
> > >>                      "mediatek,mt8135-mmc";
> > >>         reg = <0 0x11240000 0 0x1000>;
> > >>         interrupts = <GIC_SPI 72 IRQ_TYPE_LEVEL_LOW>;
> > >>         clocks = <&pericfg CLK_PERI_MSDC30_1>,
> > >>                  <&clk_null>;
> > >>         clock-names = "source", "hclk";
> > >>         status = "disabled";
> > >> };
> > >
> > > This is another case than the one we discussed about. In the case above
> > > I motivated using a dummy clock since the clock exists in the system,
> > > but is not software controllable. To abstract this from the driver
> > > (which needs this clock since it exists) we here have the dummy clock.
> > > However, of course I can't prove the clock is indeed not software
> > > controllable; that's only the information I have.
> > 
> > I was trying to answer your question "What's the motivation to send
> > this patch anyway?".
> > The motivation is to send follow on patches that use the clk_null
> > phandle.  We need to provide some clock as the mmc1's hclk.  I do not
> > understand why this has to be "clk_null", though.  It seems like this
> > should be a real clock coming from one of the real clock_controller
> > nodes.  After all, the mmc driver is going to be enabling/disabling
> > this clock for power savings at runtime.  What does that even mean for
> > clk_null ?
> 
> The original purpose of this patch is to provide a common dummy clock
> for both software don't care clock and clock that is not software
> controllable.I got comments that device tree should describe hardware
> and should put exact clock in device tree. I think this is true. So we
> will remove this clock_null patch, and:
> 
> 1. For Mediatek SoC CCF driver, James will clarify clock usage further.
> Actually, we still think it's not necessary to describe whole tree that
> software don't care, James will deal this in clock driver.

I think that aswell since the device tree is not affected in this case.
Should we realize later that we indeed need the missing clocks we can
still implement them without modifying the device tree.

> 
> 2. For other module that use SW not controllable clock (mmc case
> mentioned by Dan), because this is a real clock, we will put a dummy
> clock in device tree, like
> 
> clk_mmchclk: dummyhclk {
> 	compatible = "fixed-clock";
> 	clock-frequency = <0>;
> 	#clock-cells = <0>;
> };
> 
> How about this modification ?

I wouldn't name it 'dummy', this will again raise some eyebrows.

Sascha

-- 
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

[PATCH] arm64: dts: mt8173: add clock_null

[Eddie Huang]

Hi all,

On Wed, 2015-07-08 at 13:44 +0800, Sascha Hauer wrote:
> On Wed, Jul 08, 2015 at 10:37:21AM +0800, Eddie Huang wrote:
> > On Tue, 2015-07-07 at 23:10 +0800, Daniel Kurtz wrote:
> > > On Tue, Jul 7, 2015 at 10:36 PM, Sascha Hauer <s.hauer@pengutronix.de> wrote:
> > > > On Tue, Jul 07, 2015 at 10:15:29PM +0800, Daniel Kurtz wrote:
> > > >> On Tue, Jul 7, 2015 at 9:07 PM, Sascha Hauer <s.hauer@pengutronix.de> wrote:
> > > >> > On Thu, Jun 18, 2015 at 01:29:11PM +0800, Eddie Huang wrote:
> > > >> >> Add clk_null, which represents clocks that can not / need not
> > > >> >> controlled by software.
> > > >> >> There are many clocks' parent set to clk_null.
> > > >> >>
> > > >> >> Signed-off-by: James Liao <jamesjj.liao@mediatek.com>
> > > >> >> Signed-off-by: Eddie Huang <eddie.huang@mediatek.com>
> > > >> >> ---
> > > >> >> Base on 4.1-rc1
> > > >> >>
> > > >> >> Change-Id: I4db9b40d07e28f54f7bae9b676316cbd6a962124
> > > >> >> ---
> > > >> >>  arch/arm64/boot/dts/mediatek/mt8173.dtsi | 6 ++++++
> > > >> >>  1 file changed, 6 insertions(+)
> > > >> >>
> > > >> >> diff --git a/arch/arm64/boot/dts/mediatek/mt8173.dtsi b/arch/arm64/boot/dts/mediatek/mt8173.dtsi
> > > >> >> index 924fdb6..4798f44 100644
> > > >> >> --- a/arch/arm64/boot/dts/mediatek/mt8173.dtsi
> > > >> >> +++ b/arch/arm64/boot/dts/mediatek/mt8173.dtsi
> > > >> >> @@ -81,6 +81,12 @@
> > > >> >>               cpu_on        = <0x84000003>;
> > > >> >>       };
> > > >> >>
> > > >> >> +     clk_null: clk_null {
> > > >> >> +             compatible = "fixed-clock";
> > > >> >> +             clock-frequency = <0>;
> > > >> >> +             #clock-cells = <0>;
> > > >> >> +     };
> > > >> >
> > > >> > The discussion around this patch shows that we don't want to have this
> > > >> > clock in the device tree as it is not a hardware description.
> > > >> >
> > > >> > Ok, fine. Eddie, you told us that the rate of the current clk_null children
> > > >> > is not interesting. What's the motivation to send this patch anyway
> > > >> > then? Why can't you keep its children on the orphan list where they are
> > > >> > already now?
> > > >> >
> > > >> > Another possibility would be to instantiate the clk_null clock from C
> > > >> > code rather than from the device tree. This way we wouldn't put any
> > > >> > wrong descriptions into the device tree and still can implement the
> > > >> > support for the real parent clocks when we actually need them.
> > > >>
> > > >> Some device nodes, like mmc, use a clk_null phandle as one of their clocks:
> > > >>
> > > >> mmc1: mmc at 11240000 {
> > > >>         compatible = "mediatek,mt8173-mmc",
> > > >>                      "mediatek,mt8135-mmc";
> > > >>         reg = <0 0x11240000 0 0x1000>;
> > > >>         interrupts = <GIC_SPI 72 IRQ_TYPE_LEVEL_LOW>;
> > > >>         clocks = <&pericfg CLK_PERI_MSDC30_1>,
> > > >>                  <&clk_null>;
> > > >>         clock-names = "source", "hclk";
> > > >>         status = "disabled";
> > > >> };
> > > >
> > > > This is another case than the one we discussed about. In the case above
> > > > I motivated using a dummy clock since the clock exists in the system,
> > > > but is not software controllable. To abstract this from the driver
> > > > (which needs this clock since it exists) we here have the dummy clock.
> > > > However, of course I can't prove the clock is indeed not software
> > > > controllable; that's only the information I have.
> > > 
> > > I was trying to answer your question "What's the motivation to send
> > > this patch anyway?".
> > > The motivation is to send follow on patches that use the clk_null
> > > phandle.  We need to provide some clock as the mmc1's hclk.  I do not
> > > understand why this has to be "clk_null", though.  It seems like this
> > > should be a real clock coming from one of the real clock_controller
> > > nodes.  After all, the mmc driver is going to be enabling/disabling
> > > this clock for power savings at runtime.  What does that even mean for
> > > clk_null ?
> > 
> > The original purpose of this patch is to provide a common dummy clock
> > for both software don't care clock and clock that is not software
> > controllable.I got comments that device tree should describe hardware
> > and should put exact clock in device tree. I think this is true. So we
> > will remove this clock_null patch, and:
> > 
> > 1. For Mediatek SoC CCF driver, James will clarify clock usage further.
> > Actually, we still think it's not necessary to describe whole tree that
> > software don't care, James will deal this in clock driver.
> 
> I think that aswell since the device tree is not affected in this case.
> Should we realize later that we indeed need the missing clocks we can
> still implement them without modifying the device tree.
> 
> > 
> > 2. For other module that use SW not controllable clock (mmc case
> > mentioned by Dan), because this is a real clock, we will put a dummy
> > clock in device tree, like
> > 
> > clk_mmchclk: dummyhclk {
> > 	compatible = "fixed-clock";
> > 	clock-frequency = <0>;
> > 	#clock-cells = <0>;
> > };
> > 
> > How about this modification ?
> 
> I wouldn't name it 'dummy', this will again raise some eyebrows.
> 

I got mmc hclk from our designer. HCLK is from AXI Bus directly (sorry,
I gave wrong information to Dan and Sascha yesterday). Because there is
no any mux or gate register to control this HCLK, so current
clk-mt8173.c didn't model it. Since I know where this clock comes from,
I will abandon this stupid dummy clock device tree patch. But there are
two alternative ways:

1. In MMC device tree, use parent clock, like
     <&topckgen CLK_TOP_AXI_SEL>

2. In clk-mt8173.c, add fix factor clock, like apll case
   FACTOR(CLK_TOP_APLL1, "apll1_ck", "apll1", 1, 1),

Either way works, but have different meaning. Any suggestion to handle
thing like this.

Thanks
Eddie

[PATCH] arm64: dts: mt8173: add clock_null

[Daniel Kurtz]

On Fri, Jul 10, 2015 at 3:27 PM, Eddie Huang <eddie.huang@mediatek.com> wrote:
> Hi all,
>
> On Wed, 2015-07-08 at 13:44 +0800, Sascha Hauer wrote:
>> On Wed, Jul 08, 2015 at 10:37:21AM +0800, Eddie Huang wrote:
>> > On Tue, 2015-07-07 at 23:10 +0800, Daniel Kurtz wrote:
>> > > On Tue, Jul 7, 2015 at 10:36 PM, Sascha Hauer <s.hauer@pengutronix.de> wrote:
>> > > > On Tue, Jul 07, 2015 at 10:15:29PM +0800, Daniel Kurtz wrote:
>> > > >> On Tue, Jul 7, 2015 at 9:07 PM, Sascha Hauer <s.hauer@pengutronix.de> wrote:
>> > > >> > On Thu, Jun 18, 2015 at 01:29:11PM +0800, Eddie Huang wrote:
>> > > >> >> Add clk_null, which represents clocks that can not / need not
>> > > >> >> controlled by software.
>> > > >> >> There are many clocks' parent set to clk_null.
>> > > >> >>
>> > > >> >> Signed-off-by: James Liao <jamesjj.liao@mediatek.com>
>> > > >> >> Signed-off-by: Eddie Huang <eddie.huang@mediatek.com>
>> > > >> >> ---
>> > > >> >> Base on 4.1-rc1
>> > > >> >>
>> > > >> >> Change-Id: I4db9b40d07e28f54f7bae9b676316cbd6a962124
>> > > >> >> ---
>> > > >> >>  arch/arm64/boot/dts/mediatek/mt8173.dtsi | 6 ++++++
>> > > >> >>  1 file changed, 6 insertions(+)
>> > > >> >>
>> > > >> >> diff --git a/arch/arm64/boot/dts/mediatek/mt8173.dtsi b/arch/arm64/boot/dts/mediatek/mt8173.dtsi
>> > > >> >> index 924fdb6..4798f44 100644
>> > > >> >> --- a/arch/arm64/boot/dts/mediatek/mt8173.dtsi
>> > > >> >> +++ b/arch/arm64/boot/dts/mediatek/mt8173.dtsi
>> > > >> >> @@ -81,6 +81,12 @@
>> > > >> >>               cpu_on        = <0x84000003>;
>> > > >> >>       };
>> > > >> >>
>> > > >> >> +     clk_null: clk_null {
>> > > >> >> +             compatible = "fixed-clock";
>> > > >> >> +             clock-frequency = <0>;
>> > > >> >> +             #clock-cells = <0>;
>> > > >> >> +     };
>> > > >> >
>> > > >> > The discussion around this patch shows that we don't want to have this
>> > > >> > clock in the device tree as it is not a hardware description.
>> > > >> >
>> > > >> > Ok, fine. Eddie, you told us that the rate of the current clk_null children
>> > > >> > is not interesting. What's the motivation to send this patch anyway
>> > > >> > then? Why can't you keep its children on the orphan list where they are
>> > > >> > already now?
>> > > >> >
>> > > >> > Another possibility would be to instantiate the clk_null clock from C
>> > > >> > code rather than from the device tree. This way we wouldn't put any
>> > > >> > wrong descriptions into the device tree and still can implement the
>> > > >> > support for the real parent clocks when we actually need them.
>> > > >>
>> > > >> Some device nodes, like mmc, use a clk_null phandle as one of their clocks:
>> > > >>
>> > > >> mmc1: mmc at 11240000 {
>> > > >>         compatible = "mediatek,mt8173-mmc",
>> > > >>                      "mediatek,mt8135-mmc";
>> > > >>         reg = <0 0x11240000 0 0x1000>;
>> > > >>         interrupts = <GIC_SPI 72 IRQ_TYPE_LEVEL_LOW>;
>> > > >>         clocks = <&pericfg CLK_PERI_MSDC30_1>,
>> > > >>                  <&clk_null>;
>> > > >>         clock-names = "source", "hclk";
>> > > >>         status = "disabled";
>> > > >> };
>> > > >
>> > > > This is another case than the one we discussed about. In the case above
>> > > > I motivated using a dummy clock since the clock exists in the system,
>> > > > but is not software controllable. To abstract this from the driver
>> > > > (which needs this clock since it exists) we here have the dummy clock.
>> > > > However, of course I can't prove the clock is indeed not software
>> > > > controllable; that's only the information I have.
>> > >
>> > > I was trying to answer your question "What's the motivation to send
>> > > this patch anyway?".
>> > > The motivation is to send follow on patches that use the clk_null
>> > > phandle.  We need to provide some clock as the mmc1's hclk.  I do not
>> > > understand why this has to be "clk_null", though.  It seems like this
>> > > should be a real clock coming from one of the real clock_controller
>> > > nodes.  After all, the mmc driver is going to be enabling/disabling
>> > > this clock for power savings at runtime.  What does that even mean for
>> > > clk_null ?
>> >
>> > The original purpose of this patch is to provide a common dummy clock
>> > for both software don't care clock and clock that is not software
>> > controllable.I got comments that device tree should describe hardware
>> > and should put exact clock in device tree. I think this is true. So we
>> > will remove this clock_null patch, and:
>> >
>> > 1. For Mediatek SoC CCF driver, James will clarify clock usage further.
>> > Actually, we still think it's not necessary to describe whole tree that
>> > software don't care, James will deal this in clock driver.
>>
>> I think that aswell since the device tree is not affected in this case.
>> Should we realize later that we indeed need the missing clocks we can
>> still implement them without modifying the device tree.
>>
>> >
>> > 2. For other module that use SW not controllable clock (mmc case
>> > mentioned by Dan), because this is a real clock, we will put a dummy
>> > clock in device tree, like
>> >
>> > clk_mmchclk: dummyhclk {
>> >     compatible = "fixed-clock";
>> >     clock-frequency = <0>;
>> >     #clock-cells = <0>;
>> > };
>> >
>> > How about this modification ?
>>
>> I wouldn't name it 'dummy', this will again raise some eyebrows.
>>
>
> I got mmc hclk from our designer. HCLK is from AXI Bus directly (sorry,
> I gave wrong information to Dan and Sascha yesterday). Because there is
> no any mux or gate register to control this HCLK, so current
> clk-mt8173.c didn't model it. Since I know where this clock comes from,
> I will abandon this stupid dummy clock device tree patch. But there are
> two alternative ways:
>
> 1. In MMC device tree, use parent clock, like
>      <&topckgen CLK_TOP_AXI_SEL>
>
> 2. In clk-mt8173.c, add fix factor clock, like apll case
>    FACTOR(CLK_TOP_APLL1, "apll1_ck", "apll1", 1, 1),
>
> Either way works, but have different meaning. Any suggestion to handle
> thing like this.

I like (1), it seems more straightforward.
What is the advantage of (2)?

Thanks,
-Dan

> Thanks
> Eddie
>
>

[PATCH] arm64: dts: mt8173: add clock_null

[Eddie Huang]

On Fri, 2015-07-10 at 16:11 +0800, Daniel Kurtz wrote:
> On Fri, Jul 10, 2015 at 3:27 PM, Eddie Huang <eddie.huang@mediatek.com> wrote:
> > Hi all,
> >
> > On Wed, 2015-07-08 at 13:44 +0800, Sascha Hauer wrote:
> >> On Wed, Jul 08, 2015 at 10:37:21AM +0800, Eddie Huang wrote:
> >> > On Tue, 2015-07-07 at 23:10 +0800, Daniel Kurtz wrote:
> >> > > On Tue, Jul 7, 2015 at 10:36 PM, Sascha Hauer <s.hauer@pengutronix.de> wrote:
> >> > > > On Tue, Jul 07, 2015 at 10:15:29PM +0800, Daniel Kurtz wrote:
> >> > > >> On Tue, Jul 7, 2015 at 9:07 PM, Sascha Hauer <s.hauer@pengutronix.de> wrote:
> >> > > >> > On Thu, Jun 18, 2015 at 01:29:11PM +0800, Eddie Huang wrote:
> >> > > >> >> Add clk_null, which represents clocks that can not / need not
> >> > > >> >> controlled by software.
> >> > > >> >> There are many clocks' parent set to clk_null.
> >> > > >> >>
> >> > > >> >> Signed-off-by: James Liao <jamesjj.liao@mediatek.com>
> >> > > >> >> Signed-off-by: Eddie Huang <eddie.huang@mediatek.com>
> >> > > >> >> ---
> >> > > >> >> Base on 4.1-rc1
> >> > > >> >>
> >> > > >> >> Change-Id: I4db9b40d07e28f54f7bae9b676316cbd6a962124
> >> > > >> >> ---
> >> > > >> >>  arch/arm64/boot/dts/mediatek/mt8173.dtsi | 6 ++++++
> >> > > >> >>  1 file changed, 6 insertions(+)
> >> > > >> >>
> >> > > >> >> diff --git a/arch/arm64/boot/dts/mediatek/mt8173.dtsi b/arch/arm64/boot/dts/mediatek/mt8173.dtsi
> >> > > >> >> index 924fdb6..4798f44 100644
> >> > > >> >> --- a/arch/arm64/boot/dts/mediatek/mt8173.dtsi
> >> > > >> >> +++ b/arch/arm64/boot/dts/mediatek/mt8173.dtsi
> >> > > >> >> @@ -81,6 +81,12 @@
> >> > > >> >>               cpu_on        = <0x84000003>;
> >> > > >> >>       };
> >> > > >> >>
> >> > > >> >> +     clk_null: clk_null {
> >> > > >> >> +             compatible = "fixed-clock";
> >> > > >> >> +             clock-frequency = <0>;
> >> > > >> >> +             #clock-cells = <0>;
> >> > > >> >> +     };
> >> > > >> >
> >> > > >> > The discussion around this patch shows that we don't want to have this
> >> > > >> > clock in the device tree as it is not a hardware description.
> >> > > >> >
> >> > > >> > Ok, fine. Eddie, you told us that the rate of the current clk_null children
> >> > > >> > is not interesting. What's the motivation to send this patch anyway
> >> > > >> > then? Why can't you keep its children on the orphan list where they are
> >> > > >> > already now?
> >> > > >> >
> >> > > >> > Another possibility would be to instantiate the clk_null clock from C
> >> > > >> > code rather than from the device tree. This way we wouldn't put any
> >> > > >> > wrong descriptions into the device tree and still can implement the
> >> > > >> > support for the real parent clocks when we actually need them.
> >> > > >>
> >> > > >> Some device nodes, like mmc, use a clk_null phandle as one of their clocks:
> >> > > >>
> >> > > >> mmc1: mmc at 11240000 {
> >> > > >>         compatible = "mediatek,mt8173-mmc",
> >> > > >>                      "mediatek,mt8135-mmc";
> >> > > >>         reg = <0 0x11240000 0 0x1000>;
> >> > > >>         interrupts = <GIC_SPI 72 IRQ_TYPE_LEVEL_LOW>;
> >> > > >>         clocks = <&pericfg CLK_PERI_MSDC30_1>,
> >> > > >>                  <&clk_null>;
> >> > > >>         clock-names = "source", "hclk";
> >> > > >>         status = "disabled";
> >> > > >> };
> >> > > >
> >> > > > This is another case than the one we discussed about. In the case above
> >> > > > I motivated using a dummy clock since the clock exists in the system,
> >> > > > but is not software controllable. To abstract this from the driver
> >> > > > (which needs this clock since it exists) we here have the dummy clock.
> >> > > > However, of course I can't prove the clock is indeed not software
> >> > > > controllable; that's only the information I have.
> >> > >
> >> > > I was trying to answer your question "What's the motivation to send
> >> > > this patch anyway?".
> >> > > The motivation is to send follow on patches that use the clk_null
> >> > > phandle.  We need to provide some clock as the mmc1's hclk.  I do not
> >> > > understand why this has to be "clk_null", though.  It seems like this
> >> > > should be a real clock coming from one of the real clock_controller
> >> > > nodes.  After all, the mmc driver is going to be enabling/disabling
> >> > > this clock for power savings at runtime.  What does that even mean for
> >> > > clk_null ?
> >> >
> >> > The original purpose of this patch is to provide a common dummy clock
> >> > for both software don't care clock and clock that is not software
> >> > controllable.I got comments that device tree should describe hardware
> >> > and should put exact clock in device tree. I think this is true. So we
> >> > will remove this clock_null patch, and:
> >> >
> >> > 1. For Mediatek SoC CCF driver, James will clarify clock usage further.
> >> > Actually, we still think it's not necessary to describe whole tree that
> >> > software don't care, James will deal this in clock driver.
> >>
> >> I think that aswell since the device tree is not affected in this case.
> >> Should we realize later that we indeed need the missing clocks we can
> >> still implement them without modifying the device tree.
> >>
> >> >
> >> > 2. For other module that use SW not controllable clock (mmc case
> >> > mentioned by Dan), because this is a real clock, we will put a dummy
> >> > clock in device tree, like
> >> >
> >> > clk_mmchclk: dummyhclk {
> >> >     compatible = "fixed-clock";
> >> >     clock-frequency = <0>;
> >> >     #clock-cells = <0>;
> >> > };
> >> >
> >> > How about this modification ?
> >>
> >> I wouldn't name it 'dummy', this will again raise some eyebrows.
> >>
> >
> > I got mmc hclk from our designer. HCLK is from AXI Bus directly (sorry,
> > I gave wrong information to Dan and Sascha yesterday). Because there is
> > no any mux or gate register to control this HCLK, so current
> > clk-mt8173.c didn't model it. Since I know where this clock comes from,
> > I will abandon this stupid dummy clock device tree patch. But there are
> > two alternative ways:
> >
> > 1. In MMC device tree, use parent clock, like
> >      <&topckgen CLK_TOP_AXI_SEL>
> >
> > 2. In clk-mt8173.c, add fix factor clock, like apll case
> >    FACTOR(CLK_TOP_APLL1, "apll1_ck", "apll1", 1, 1),
> >
> > Either way works, but have different meaning. Any suggestion to handle
> > thing like this.
> 
> I like (1), it seems more straightforward.
> What is the advantage of (2)?
> 

Actually no, it can't even guarantee HCLK parent clock reference count
neither. We will resend mmc device tree patch base on this solution.

Thanks
Eddie