diff mbox

[3/3] ARM: imx6: Fix procedure to switch the parent of LDB_DI_CLK

Message ID 1420559265-7333-3-git-send-email-p.zabel@pengutronix.de (mailing list archive)
State New, archived
Headers show

Commit Message

Philipp Zabel Jan. 6, 2015, 3:47 p.m. UTC
From: Fabio Estevam <fabio.estevam@freescale.com>

Due to incorrect placement of the clock gate cell in the ldb_di[x]_clk tree,
the glitchy parent mux of ldb_di[x]_clk can cause a glitch to enter the
ldb_di_ipu_div divider. If the divider gets locked up, no ldb_di[x]_clk is
generated, and the LVDS display will hang when the ipu_di_clk is sourced from
ldb_di_clk.

To fix the problem, both the new and current parent of the ldb_di_clk should
be disabled before the switch. This patch ensures that correct steps are
followed when ldb_di_clk parent is switched in the beginning of boot. The
glitchy muxes are then registered as read-only. The clock parent can be selected
using the assigned-clocks and assigned-clock-parents properties of the ccm
device tree node:

	&clks {
		assigned-clocks = <&clks IMX6QDL_CLK_LDB_DI0_SEL>,
				  <&clks IMX6QDL_CLK_LDB_DI1_SEL>;
		assigned-clock-parents = <&clks IMX6QDL_CLK_MMDC_CH1_AXI>,
					 <&clks IMX6QDL_CLK_PLL5_VIDEO_DIV>;
	};

Signed-off-by: Ranjani Vaidyanathan <Ranjani.Vaidyanathan@freescale.com>
Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
---
 arch/arm/mach-imx/clk-imx6q.c | 201 ++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 201 insertions(+)

Comments

Shawn Guo Jan. 8, 2015, 1 p.m. UTC | #1
On Tue, Jan 06, 2015 at 04:47:45PM +0100, Philipp Zabel wrote:
> From: Fabio Estevam <fabio.estevam@freescale.com>
> 
> Due to incorrect placement of the clock gate cell in the ldb_di[x]_clk tree,
> the glitchy parent mux of ldb_di[x]_clk can cause a glitch to enter the
> ldb_di_ipu_div divider. If the divider gets locked up, no ldb_di[x]_clk is
> generated, and the LVDS display will hang when the ipu_di_clk is sourced from
> ldb_di_clk.
> 
> To fix the problem, both the new and current parent of the ldb_di_clk should
> be disabled before the switch. This patch ensures that correct steps are
> followed when ldb_di_clk parent is switched in the beginning of boot. The
> glitchy muxes are then registered as read-only. The clock parent can be selected
> using the assigned-clocks and assigned-clock-parents properties of the ccm
> device tree node:
> 
> 	&clks {
> 		assigned-clocks = <&clks IMX6QDL_CLK_LDB_DI0_SEL>,
> 				  <&clks IMX6QDL_CLK_LDB_DI1_SEL>;
> 		assigned-clock-parents = <&clks IMX6QDL_CLK_MMDC_CH1_AXI>,
> 					 <&clks IMX6QDL_CLK_PLL5_VIDEO_DIV>;
> 	};
> 
> Signed-off-by: Ranjani Vaidyanathan <Ranjani.Vaidyanathan@freescale.com>
> Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
> Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>

Before I start reviewing the patch, I would like to know if there is a
formal errata for this issue, including the bug description, recommended
workaround, and affected i.MX6 variants etc.

Shawn
Fabio Estevam Jan. 15, 2015, 4:54 p.m. UTC | #2
Hi Shawn,

On Thu, Jan 8, 2015 at 11:00 AM, Shawn Guo <shawn.guo@linaro.org> wrote:

> Before I start reviewing the patch, I would like to know if there is a
> formal errata for this issue, including the bug description, recommended
> workaround, and affected i.MX6 variants etc.

The latest mx6 reference manual available on the web shows the LDB
clock gate in the correct position now.

It also adds the following note:

"NOTE: This clock path is not controlled by a clock gate (CG). For
proper clock switching procedures refer to the Clock Switching
Multiplexers section."

There is no formal errata available.

Please consider reviewing this series, as it addresses a real problem
that affect many users.

Thanks,

Fabio Estevam
Dirk Behme Jan. 16, 2015, 9:28 a.m. UTC | #3
On 15.01.2015 17:54, Fabio Estevam wrote:
> Hi Shawn,
>
> On Thu, Jan 8, 2015 at 11:00 AM, Shawn Guo <shawn.guo@linaro.org> wrote:
>
>> Before I start reviewing the patch, I would like to know if there is a
>> formal errata for this issue, including the bug description, recommended
>> workaround, and affected i.MX6 variants etc.
>
> The latest mx6 reference manual available on the web shows the LDB
> clock gate in the correct position now.
>
> It also adds the following note:
>
> "NOTE: This clock path is not controlled by a clock gate (CG). For
> proper clock switching procedures refer to the Clock Switching
> Multiplexers section."
>
> There is no formal errata available.
>
> Please consider reviewing this series, as it addresses a real problem
> that affect many users.


Just to get a better understanding for the background of this series: 
Are the patches in this series somehow related to Shawn's ENGR00318063-x 
patches in the FSL 3.10.53_1.1.0_ga kernel:

http://git.freescale.com/git/cgit.cgi/imx/linux-2.6-imx.git/log/?h=imx_3.10.53_1.1.0_ga&qt=grep&q=ENGR00318063

?

Or are this completely different topics?

Best regards

Dirk
Shawn Guo Jan. 20, 2015, 6 a.m. UTC | #4
On Thu, Jan 15, 2015 at 02:54:24PM -0200, Fabio Estevam wrote:
> Hi Shawn,
> 
> On Thu, Jan 8, 2015 at 11:00 AM, Shawn Guo <shawn.guo@linaro.org> wrote:
> 
> > Before I start reviewing the patch, I would like to know if there is a
> > formal errata for this issue, including the bug description, recommended
> > workaround, and affected i.MX6 variants etc.
> 
> The latest mx6 reference manual available on the web shows the LDB
> clock gate in the correct position now.
> 
> It also adds the following note:
> 
> "NOTE: This clock path is not controlled by a clock gate (CG). For
> proper clock switching procedures refer to the Clock Switching
> Multiplexers section."

No.  The Clock Switching Multiplexers section documents the procedure
for the glitchy multiplexer with a clock gate in the right place, not
this buggy LDB multiplexer.

> 
> There is no formal errata available.
> 
> Please consider reviewing this series, as it addresses a real problem
> that affect many users.

I'm not going to review/take the patch until I see a formal errata
documenting the software workaround procedure for the issue.

Shawn
Shawn Guo Jan. 20, 2015, 6:13 a.m. UTC | #5
On Fri, Jan 16, 2015 at 10:28:46AM +0100, Dirk Behme wrote:
> On 15.01.2015 17:54, Fabio Estevam wrote:
> >Hi Shawn,
> >
> >On Thu, Jan 8, 2015 at 11:00 AM, Shawn Guo <shawn.guo@linaro.org> wrote:
> >
> >>Before I start reviewing the patch, I would like to know if there is a
> >>formal errata for this issue, including the bug description, recommended
> >>workaround, and affected i.MX6 variants etc.
> >
> >The latest mx6 reference manual available on the web shows the LDB
> >clock gate in the correct position now.
> >
> >It also adds the following note:
> >
> >"NOTE: This clock path is not controlled by a clock gate (CG). For
> >proper clock switching procedures refer to the Clock Switching
> >Multiplexers section."
> >
> >There is no formal errata available.
> >
> >Please consider reviewing this series, as it addresses a real problem
> >that affect many users.
> 
> 
> Just to get a better understanding for the background of this
> series: Are the patches in this series somehow related to Shawn's
> ENGR00318063-x patches in the FSL 3.10.53_1.1.0_ga kernel:
> 
> http://git.freescale.com/git/cgit.cgi/imx/linux-2.6-imx.git/log/?h=imx_3.10.53_1.1.0_ga&qt=grep&q=ENGR00318063
> 
> ?

Fabio,

Look, without the errata for this problem, your customer do not even
know what this patch is for.


Dirk,

Yes, it tries to fix the buggy ldb_di_ipu clock output which doesn't
have a gate between ldb_di_clk_sel and div_3.5_7.  That's also part of
the issues that ENGR00318063-x patches try to address.

Shawn
Dirk Behme Feb. 10, 2015, 11:24 a.m. UTC | #6
On 06.01.2015 16:47, Philipp Zabel wrote:
> From: Fabio Estevam <fabio.estevam@freescale.com>
>
> Due to incorrect placement of the clock gate cell in the ldb_di[x]_clk tree,
> the glitchy parent mux of ldb_di[x]_clk can cause a glitch to enter the
> ldb_di_ipu_div divider. If the divider gets locked up, no ldb_di[x]_clk is
> generated, and the LVDS display will hang when the ipu_di_clk is sourced from
> ldb_di_clk.
>
> To fix the problem, both the new and current parent of the ldb_di_clk should
> be disabled before the switch. This patch ensures that correct steps are
> followed when ldb_di_clk parent is switched in the beginning of boot. The
> glitchy muxes are then registered as read-only. The clock parent can be selected
> using the assigned-clocks and assigned-clock-parents properties of the ccm
> device tree node:
>
> 	&clks {
> 		assigned-clocks = <&clks IMX6QDL_CLK_LDB_DI0_SEL>,
> 				  <&clks IMX6QDL_CLK_LDB_DI1_SEL>;
> 		assigned-clock-parents = <&clks IMX6QDL_CLK_MMDC_CH1_AXI>,
> 					 <&clks IMX6QDL_CLK_PLL5_VIDEO_DIV>;
> 	};
>
> Signed-off-by: Ranjani Vaidyanathan <Ranjani.Vaidyanathan@freescale.com>
> Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
> Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
> ---
>   arch/arm/mach-imx/clk-imx6q.c | 201 ++++++++++++++++++++++++++++++++++++++++++
>   1 file changed, 201 insertions(+)
>
> diff --git a/arch/arm/mach-imx/clk-imx6q.c b/arch/arm/mach-imx/clk-imx6q.c
> index 2e379fb..77aa7b17 100644
> --- a/arch/arm/mach-imx/clk-imx6q.c
> +++ b/arch/arm/mach-imx/clk-imx6q.c
...
> +	/*
> +	 * The LDB_DI0/1_SEL muxes are registered read-only due to a hardware
> +	 * bug. Set the muxes to the requested values before registering the
> +	 * ldb_di_sel clocks.
> +	 */
> +	if ((imx_get_soc_revision() != IMX_CHIP_REVISION_1_0) ||
> +	    cpu_is_imx6dl())
> +		init_ldb_clks(np, base);


I've a hopefully simple understanding question regarding this if():

What does happen (or: has to be done) in the (non-existing) else case of 
this if() statement?

E.g. how to configure the clocks for the imx_get_soc_revision() == 
IMX_CHIP_REVISION_1_0 case?

For the if() case, we can configure the clocks via the assigned-clocks 
and assigned-clock-parents as described in the commit message. But how 
to configure the clocks/muxes in the else case? To my understanding, 
with the patch 2/3 [1] the clocks/muxes are configured to read only.

Sorry if I missed anything ;)

Many thanks,

Dirk

[1] 
http://lists.infradead.org/pipermail/linux-arm-kernel/2015-January/313619.html
Joshua Clayton Feb. 10, 2015, 3:31 p.m. UTC | #7
On Tuesday, February 10, 2015 12:24:29 PM Dirk Behme wrote:
> On 06.01.2015 16:47, Philipp Zabel wrote:
> > From: Fabio Estevam <fabio.estevam@freescale.com>
> > 
> > Due to incorrect placement of the clock gate cell in the ldb_di[x]_clk
> > tree, the glitchy parent mux of ldb_di[x]_clk can cause a glitch to enter
> > the ldb_di_ipu_div divider. If the divider gets locked up, no
> > ldb_di[x]_clk is generated, and the LVDS display will hang when the
> > ipu_di_clk is sourced from ldb_di_clk.
> > 
> > To fix the problem, both the new and current parent of the ldb_di_clk
> > should be disabled before the switch. This patch ensures that correct
> > steps are followed when ldb_di_clk parent is switched in the beginning of
> > boot. The glitchy muxes are then registered as read-only. The clock
> > parent can be selected using the assigned-clocks and
> > assigned-clock-parents properties of the ccm> 
> > device tree node:
> > 	&clks {
> > 	
> > 		assigned-clocks = <&clks IMX6QDL_CLK_LDB_DI0_SEL>,
> > 		
> > 				  <&clks IMX6QDL_CLK_LDB_DI1_SEL>;
> > 		
> > 		assigned-clock-parents = <&clks IMX6QDL_CLK_MMDC_CH1_AXI>,
> > 		
> > 					 <&clks IMX6QDL_CLK_PLL5_VIDEO_DIV>;
> > 	
> > 	};
> > 
> > Signed-off-by: Ranjani Vaidyanathan <Ranjani.Vaidyanathan@freescale.com>
> > Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
> > Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
> > ---
> > 
> >   arch/arm/mach-imx/clk-imx6q.c | 201
> >   ++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 201
> >   insertions(+)
> > 
> > diff --git a/arch/arm/mach-imx/clk-imx6q.c b/arch/arm/mach-imx/clk-imx6q.c
> > index 2e379fb..77aa7b17 100644
> > --- a/arch/arm/mach-imx/clk-imx6q.c
> > +++ b/arch/arm/mach-imx/clk-imx6q.c
> 
> ...
> 
> > +	/*
> > +	 * The LDB_DI0/1_SEL muxes are registered read-only due to a hardware
> > +	 * bug. Set the muxes to the requested values before registering the
> > +	 * ldb_di_sel clocks.
> > +	 */
> > +	if ((imx_get_soc_revision() != IMX_CHIP_REVISION_1_0) ||
> > +	    cpu_is_imx6dl())
> > +		init_ldb_clks(np, base);
> 
> I've a hopefully simple understanding question regarding this if():
> 
> What does happen (or: has to be done) in the (non-existing) else case of
> this if() statement?
> 
> E.g. how to configure the clocks for the imx_get_soc_revision() ==
> IMX_CHIP_REVISION_1_0 case?
> 
> For the if() case, we can configure the clocks via the assigned-clocks
> and assigned-clock-parents as described in the commit message. But how
> to configure the clocks/muxes in the else case? To my understanding,
> with the patch 2/3 [1] the clocks/muxes are configured to read only.
> 
> Sorry if I missed anything ;)
> 
> Many thanks,
> 
> Dirk
> 
> [1]
> http://lists.infradead.org/pipermail/linux-arm-kernel/2015-January/313619.ht
> ml
> 
> 
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
I have at least one imx6q SoM with IMX_CHIP_REVISION_1_0, and can confirm
that with the current mainline kernel, it gets no lvds video,
or rather very little.
I see penguins, and then it goes black.

with the 3.0.35 vendor kernel it was working.
My only clue so far has been that there are a couple of clocks and
regulators that come up different on the newer chip.

vddarm and vddpu are 1100 mV on 1.0, 1150 mV on 1.5

and there is a message that looks like so on 1.0 chip:
Switching to timer-based delay loop, resolution 15ns
and like so on 1.5 (aka 1.3d):
Switching to timer-based delay loop, resolution 333ns

Hope that helps.
Dirk Behme Feb. 10, 2015, 5:21 p.m. UTC | #8
On 10.02.2015 16:31, Joshua Clayton wrote:
> On Tuesday, February 10, 2015 12:24:29 PM Dirk Behme wrote:
>> On 06.01.2015 16:47, Philipp Zabel wrote:
>>> From: Fabio Estevam <fabio.estevam@freescale.com>
>>>
>>> Due to incorrect placement of the clock gate cell in the ldb_di[x]_clk
>>> tree, the glitchy parent mux of ldb_di[x]_clk can cause a glitch to enter
>>> the ldb_di_ipu_div divider. If the divider gets locked up, no
>>> ldb_di[x]_clk is generated, and the LVDS display will hang when the
>>> ipu_di_clk is sourced from ldb_di_clk.
>>>
>>> To fix the problem, both the new and current parent of the ldb_di_clk
>>> should be disabled before the switch. This patch ensures that correct
>>> steps are followed when ldb_di_clk parent is switched in the beginning of
>>> boot. The glitchy muxes are then registered as read-only. The clock
>>> parent can be selected using the assigned-clocks and
>>> assigned-clock-parents properties of the ccm>
>>> device tree node:
>>> 	&clks {
>>> 	
>>> 		assigned-clocks = <&clks IMX6QDL_CLK_LDB_DI0_SEL>,
>>> 		
>>> 				  <&clks IMX6QDL_CLK_LDB_DI1_SEL>;
>>> 		
>>> 		assigned-clock-parents = <&clks IMX6QDL_CLK_MMDC_CH1_AXI>,
>>> 		
>>> 					 <&clks IMX6QDL_CLK_PLL5_VIDEO_DIV>;
>>> 	
>>> 	};
>>>
>>> Signed-off-by: Ranjani Vaidyanathan <Ranjani.Vaidyanathan@freescale.com>
>>> Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
>>> Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
>>> ---
>>>
>>>    arch/arm/mach-imx/clk-imx6q.c | 201
>>>    ++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 201
>>>    insertions(+)
>>>
>>> diff --git a/arch/arm/mach-imx/clk-imx6q.c b/arch/arm/mach-imx/clk-imx6q.c
>>> index 2e379fb..77aa7b17 100644
>>> --- a/arch/arm/mach-imx/clk-imx6q.c
>>> +++ b/arch/arm/mach-imx/clk-imx6q.c
>>
>> ...
>>
>>> +	/*
>>> +	 * The LDB_DI0/1_SEL muxes are registered read-only due to a hardware
>>> +	 * bug. Set the muxes to the requested values before registering the
>>> +	 * ldb_di_sel clocks.
>>> +	 */
>>> +	if ((imx_get_soc_revision() != IMX_CHIP_REVISION_1_0) ||
>>> +	    cpu_is_imx6dl())
>>> +		init_ldb_clks(np, base);
>>
>> I've a hopefully simple understanding question regarding this if():
>>
>> What does happen (or: has to be done) in the (non-existing) else case of
>> this if() statement?
>>
>> E.g. how to configure the clocks for the imx_get_soc_revision() ==
>> IMX_CHIP_REVISION_1_0 case?
>>
>> For the if() case, we can configure the clocks via the assigned-clocks
>> and assigned-clock-parents as described in the commit message. But how
>> to configure the clocks/muxes in the else case? To my understanding,
>> with the patch 2/3 [1] the clocks/muxes are configured to read only.
>>
>> Sorry if I missed anything ;)
>>
>> Many thanks,
>>
>> Dirk
>>
>> [1]
>> http://lists.infradead.org/pipermail/linux-arm-kernel/2015-January/313619.ht
>> ml
....
> I have at least one imx6q SoM with IMX_CHIP_REVISION_1_0, and can confirm
> that with the current mainline kernel, it gets no lvds video,

Just to avoid confusion: I don't think the patch we are talking about 
here is already in current mainline:

https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/log/arch/arm/mach-imx/clk-imx6q.c

I.e. the issue you describe is an other topic than the one I asked for.

Best regards

Dirk
Dirk Behme Feb. 26, 2015, 6:32 a.m. UTC | #9
On 10.02.2015 12:24, Dirk Behme wrote:
> On 06.01.2015 16:47, Philipp Zabel wrote:
>> From: Fabio Estevam <fabio.estevam@freescale.com>
>>
>> Due to incorrect placement of the clock gate cell in the ldb_di[x]_clk
>> tree,
>> the glitchy parent mux of ldb_di[x]_clk can cause a glitch to enter the
>> ldb_di_ipu_div divider. If the divider gets locked up, no
>> ldb_di[x]_clk is
>> generated, and the LVDS display will hang when the ipu_di_clk is
>> sourced from
>> ldb_di_clk.
>>
>> To fix the problem, both the new and current parent of the ldb_di_clk
>> should
>> be disabled before the switch. This patch ensures that correct steps are
>> followed when ldb_di_clk parent is switched in the beginning of boot. The
>> glitchy muxes are then registered as read-only. The clock parent can
>> be selected
>> using the assigned-clocks and assigned-clock-parents properties of the
>> ccm
>> device tree node:
>>
>>     &clks {
>>         assigned-clocks = <&clks IMX6QDL_CLK_LDB_DI0_SEL>,
>>                   <&clks IMX6QDL_CLK_LDB_DI1_SEL>;
>>         assigned-clock-parents = <&clks IMX6QDL_CLK_MMDC_CH1_AXI>,
>>                      <&clks IMX6QDL_CLK_PLL5_VIDEO_DIV>;
>>     };
>>
>> Signed-off-by: Ranjani Vaidyanathan <Ranjani.Vaidyanathan@freescale.com>
>> Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
>> Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
>> ---
>>   arch/arm/mach-imx/clk-imx6q.c | 201
>> ++++++++++++++++++++++++++++++++++++++++++
>>   1 file changed, 201 insertions(+)
>>
>> diff --git a/arch/arm/mach-imx/clk-imx6q.c
>> b/arch/arm/mach-imx/clk-imx6q.c
>> index 2e379fb..77aa7b17 100644
>> --- a/arch/arm/mach-imx/clk-imx6q.c
>> +++ b/arch/arm/mach-imx/clk-imx6q.c
> ...
>> +    /*
>> +     * The LDB_DI0/1_SEL muxes are registered read-only due to a
>> hardware
>> +     * bug. Set the muxes to the requested values before registering the
>> +     * ldb_di_sel clocks.
>> +     */
>> +    if ((imx_get_soc_revision() != IMX_CHIP_REVISION_1_0) ||
>> +        cpu_is_imx6dl())
>> +        init_ldb_clks(np, base);
>
>
> I've a hopefully simple understanding question regarding this if():
>
> What does happen (or: has to be done) in the (non-existing) else case of
> this if() statement?
>
> E.g. how to configure the clocks for the imx_get_soc_revision() ==
> IMX_CHIP_REVISION_1_0 case?
>
> For the if() case, we can configure the clocks via the assigned-clocks
> and assigned-clock-parents as described in the commit message. But how
> to configure the clocks/muxes in the else case? To my understanding,
> with the patch 2/3 [1] the clocks/muxes are configured to read only.
>
> Sorry if I missed anything ;)


Any ideas regarding this?

Many thanks

Dirk

> [1]
> http://lists.infradead.org/pipermail/linux-arm-kernel/2015-January/313619.html
Philipp Zabel Feb. 26, 2015, 8:51 a.m. UTC | #10
Hi Dirk,

Am Donnerstag, den 26.02.2015, 07:32 +0100 schrieb Dirk Behme:
> On 10.02.2015 12:24, Dirk Behme wrote:
> > On 06.01.2015 16:47, Philipp Zabel wrote:
> >> From: Fabio Estevam <fabio.estevam@freescale.com>
> >>
> >> Due to incorrect placement of the clock gate cell in the ldb_di[x]_clk
> >> tree,
> >> the glitchy parent mux of ldb_di[x]_clk can cause a glitch to enter the
> >> ldb_di_ipu_div divider. If the divider gets locked up, no
> >> ldb_di[x]_clk is
> >> generated, and the LVDS display will hang when the ipu_di_clk is
> >> sourced from
> >> ldb_di_clk.
> >>
> >> To fix the problem, both the new and current parent of the ldb_di_clk
> >> should
> >> be disabled before the switch. This patch ensures that correct steps are
> >> followed when ldb_di_clk parent is switched in the beginning of boot. The
> >> glitchy muxes are then registered as read-only. The clock parent can
> >> be selected
> >> using the assigned-clocks and assigned-clock-parents properties of the
> >> ccm
> >> device tree node:
> >>
> >>     &clks {
> >>         assigned-clocks = <&clks IMX6QDL_CLK_LDB_DI0_SEL>,
> >>                   <&clks IMX6QDL_CLK_LDB_DI1_SEL>;
> >>         assigned-clock-parents = <&clks IMX6QDL_CLK_MMDC_CH1_AXI>,
> >>                      <&clks IMX6QDL_CLK_PLL5_VIDEO_DIV>;
> >>     };
> >>
> >> Signed-off-by: Ranjani Vaidyanathan <Ranjani.Vaidyanathan@freescale.com>
> >> Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
> >> Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
> >> ---
> >>   arch/arm/mach-imx/clk-imx6q.c | 201
> >> ++++++++++++++++++++++++++++++++++++++++++
> >>   1 file changed, 201 insertions(+)
> >>
> >> diff --git a/arch/arm/mach-imx/clk-imx6q.c
> >> b/arch/arm/mach-imx/clk-imx6q.c
> >> index 2e379fb..77aa7b17 100644
> >> --- a/arch/arm/mach-imx/clk-imx6q.c
> >> +++ b/arch/arm/mach-imx/clk-imx6q.c
> > ...
> >> +    /*
> >> +     * The LDB_DI0/1_SEL muxes are registered read-only due to a
> >> hardware
> >> +     * bug. Set the muxes to the requested values before registering the
> >> +     * ldb_di_sel clocks.
> >> +     */
> >> +    if ((imx_get_soc_revision() != IMX_CHIP_REVISION_1_0) ||
> >> +        cpu_is_imx6dl())
> >> +        init_ldb_clks(np, base);
> >
> >
> > I've a hopefully simple understanding question regarding this if():
> >
> > What does happen (or: has to be done) in the (non-existing) else case of
> > this if() statement?
> >
> > E.g. how to configure the clocks for the imx_get_soc_revision() ==
> > IMX_CHIP_REVISION_1_0 case?
> >
> > For the if() case, we can configure the clocks via the assigned-clocks
> > and assigned-clock-parents as described in the commit message. But how
> > to configure the clocks/muxes in the else case? To my understanding,
> > with the patch 2/3 [1] the clocks/muxes are configured to read only.
> >
> > Sorry if I missed anything ;)
>
> Any ideas regarding this?

I guess for the ((imx_get_soc_revision() == IMX_CHIP_REVISION_1_0) && !
cpu_is_imx6dl()) case we could still call init_ldb_clks and change
ldb_di_sel_clock_id to return -ENOENT for IMX6QDL_CLK_PLL5_VIDEO_DEV.
And the default should be changed.

regards
Philipp
diff mbox

Patch

diff --git a/arch/arm/mach-imx/clk-imx6q.c b/arch/arm/mach-imx/clk-imx6q.c
index 2e379fb..77aa7b17 100644
--- a/arch/arm/mach-imx/clk-imx6q.c
+++ b/arch/arm/mach-imx/clk-imx6q.c
@@ -120,9 +120,80 @@  static unsigned int share_count_ssi1;
 static unsigned int share_count_ssi2;
 static unsigned int share_count_ssi3;
 
+static int ldb_di_sel_by_clock_id(int clock_id)
+{
+	switch (clock_id) {
+	case IMX6QDL_CLK_PLL5_VIDEO_DIV:
+		return 0;
+	case IMX6QDL_CLK_PLL2_PFD0_352M:
+		return 1;
+	case IMX6QDL_CLK_PLL2_PFD2_396M:
+		return 2;
+	case IMX6QDL_CLK_MMDC_CH1_AXI:
+		return 3;
+	case IMX6QDL_CLK_PLL3_USB_OTG:
+		return 4;
+	default:
+		return -ENOENT;
+	}
+}
+
+static void of_assigned_ldb_sels(struct device_node *node,
+				 int *ldb_di0_sel, int *ldb_di1_sel)
+{
+	struct of_phandle_args clkspec;
+	int index, rc, num_parents;
+	int parent, child, sel;
+
+	num_parents = of_count_phandle_with_args(node, "assigned-clock-parents",
+						 "#clock-cells");
+	for (index = 0; index < num_parents; index++) {
+		rc = of_parse_phandle_with_args(node, "assigned-clock-parents",
+					"#clock-cells", index, &clkspec);
+		if (rc < 0) {
+			/* skip empty (null) phandles */
+			if (rc == -ENOENT)
+				continue;
+			else
+				return;
+		}
+		if (clkspec.np != node || clkspec.args[0] >= IMX6QDL_CLK_END) {
+			pr_err("ccm: parent clock %d not in ccm\n", index);
+			return;
+		}
+		parent = clkspec.args[0];
+
+		rc = of_parse_phandle_with_args(node, "assigned-clocks",
+				"#clock-cells", index, &clkspec);
+		if (rc < 0)
+			return;
+		if (clkspec.np != node || clkspec.args[0] >= IMX6QDL_CLK_END) {
+			pr_err("ccm: child clock %d not in ccm\n", index);
+			return;
+		}
+		child = clkspec.args[0];
+
+		if (child != IMX6QDL_CLK_LDB_DI0_SEL &&
+		    child != IMX6QDL_CLK_LDB_DI1_SEL)
+			continue;
+
+		sel = ldb_di_sel_by_clock_id(parent);
+		if (sel < 0)
+			pr_err("ccm: invalid ldb_di parent clock\n");
+
+		if (child == IMX6QDL_CLK_LDB_DI0_SEL)
+			*ldb_di0_sel = sel;
+		if (child == IMX6QDL_CLK_LDB_DI1_SEL)
+			*ldb_di1_sel = sel;
+	}
+}
+
 #define CCM_CCDR		0x04
+#define CCM_CCSR		0x0c
+#define CCM_CS2CDR		0x2c
 
 #define CCDR_MMDC_CH1_MASK	BIT(16)
+#define CCSR_PLL3_SW_CLK_SEL	BIT(0)
 
 static void __init imx6q_mmdc_ch1_mask_handshake(void __iomem *ccm_base)
 {
@@ -133,6 +204,124 @@  static void __init imx6q_mmdc_ch1_mask_handshake(void __iomem *ccm_base)
 	writel_relaxed(reg, ccm_base + CCM_CCDR);
 }
 
+/*
+ * The only way to disable the MMDC_CH1 clock is to move it to pll3_sw_clk
+ * via periph2_clk2_sel and then to disable pll3_sw_clk by selecting the
+ * bypass clock source, since there is no CG bit for mmdc_ch1.
+ */
+static void mmdc_ch1_disable(void __iomem *ccm_base)
+{
+	unsigned int reg;
+
+	clk_set_parent(clk[IMX6QDL_CLK_PERIPH2_CLK2_SEL],
+		       clk[IMX6QDL_CLK_PLL3_USB_OTG]);
+
+	/*
+	 * Handshake with mmdc_ch1 module must be masked when changing
+	 * periph2_clk_sel.
+	 */
+	clk_set_parent(clk[IMX6QDL_CLK_PERIPH2], clk[IMX6QDL_CLK_PERIPH2_CLK2]);
+
+	/* Disable pll3_sw_clk by selecting the bypass clock source */
+	reg = readl_relaxed(ccm_base + CCM_CCSR);
+	reg |= CCSR_PLL3_SW_CLK_SEL;
+	writel_relaxed(reg, ccm_base + CCM_CCSR);
+}
+
+static void mmdc_ch1_reenable(void __iomem *ccm_base)
+{
+	unsigned int reg;
+
+	/* Enable pll3_sw_clk by disabling the bypass */
+	reg = readl_relaxed(ccm_base + CCM_CCSR);
+	reg &= ~CCSR_PLL3_SW_CLK_SEL;
+	writel_relaxed(reg, ccm_base + CCM_CCSR);
+
+	clk_set_parent(clk[IMX6QDL_CLK_PERIPH2], clk[IMX6QDL_CLK_PERIPH2_PRE]);
+}
+
+/*
+ * Need to follow a strict procedure when changing the LDB
+ * clock, else we can introduce a glitch. Things to keep in
+ * mind:
+ * 1. The current and new parent clocks must be disabled.
+ * 2. The default clock for ldb_dio_clk is mmdc_ch1 which has
+ * no CG bit.
+ * 3. In the RTL implementation of the LDB_DI_CLK_SEL mux
+ * the top four options are in one mux and the PLL3 option along
+ * with another option is in the second mux. There is third mux
+ * used to decide between the first and second mux.
+ * The code below switches the parent to the bottom mux first
+ * and then manipulates the top mux. This ensures that no glitch
+ * will enter the divider.
+ */
+static void init_ldb_clks(struct device_node *np, void __iomem *ccm_base)
+{
+	unsigned int reg;
+	int ldb_di0_sel[4] = { 0 };
+	int ldb_di1_sel[4] = { 0 };
+	int i;
+
+	reg = readl_relaxed(ccm_base + CCM_CS2CDR);
+	ldb_di0_sel[0] = (reg >> 9) & 7;
+	ldb_di1_sel[0] = (reg >> 12) & 7;
+
+	of_assigned_ldb_sels(np, &ldb_di0_sel[3], &ldb_di1_sel[3]);
+
+	if (ldb_di0_sel[0] == ldb_di0_sel[3] &&
+	    ldb_di1_sel[0] == ldb_di1_sel[3])
+		return;
+
+	if (ldb_di0_sel[0] != 3 || ldb_di1_sel[0] != 3)
+		pr_warn("ccm: ldb_di_sel already changed from reset value\n");
+
+	if (ldb_di0_sel[0] > 3 || ldb_di1_sel[0] > 3 ||
+	    ldb_di0_sel[3] > 3 || ldb_di1_sel[3] > 3) {
+		pr_err("ccm: ldb_di_sel workaround only for top mux\n");
+		return;
+	}
+
+	ldb_di0_sel[1] = ldb_di0_sel[0] | 4;
+	ldb_di0_sel[2] = ldb_di0_sel[3] | 4;
+	ldb_di1_sel[1] = ldb_di1_sel[0] | 4;
+	ldb_di1_sel[2] = ldb_di1_sel[3] | 4;
+
+	mmdc_ch1_disable(ccm_base);
+
+	for (i = 1; i < 4; i++) {
+		reg = readl_relaxed(ccm_base + CCM_CS2CDR);
+		reg &= ~((7 << 9) | (7 << 12));
+		reg |= ((ldb_di0_sel[i] << 9) | (ldb_di1_sel[i] << 12));
+		writel_relaxed(reg, ccm_base + CCM_CS2CDR);
+	}
+
+	mmdc_ch1_reenable(ccm_base);
+}
+
+static void disable_anatop_clocks(void __iomem *anatop_base)
+{
+	unsigned int reg;
+
+	/* Make sure PFDs are disabled at boot. */
+	reg = readl_relaxed(anatop_base + 0x100);
+	/* Cannot gate PFD2 if pll2_pfd2_396m is the parent of MMDC clock */
+	if (clk_get_parent(clk[IMX6QDL_CLK_PERIPH_PRE]) ==
+	    clk[IMX6QDL_CLK_PLL2_PFD2_396M])
+		reg |= 0x00008080;
+	else
+		reg |= 0x00808080;
+	writel_relaxed(reg, anatop_base + 0x100);
+
+	reg = readl_relaxed(anatop_base + 0xf0);
+	reg |= 0x80808080;
+	writel_relaxed(reg, anatop_base + 0xf0);
+
+	/* Make sure PLLs is disabled */
+	reg = readl_relaxed(anatop_base + 0xa0);
+	reg &= ~(1 << 13);
+	writel_relaxed(reg, anatop_base + 0xa0);
+}
+
 static void __init imx6q_clocks_init(struct device_node *ccm_node)
 {
 	struct device_node *np;
@@ -269,6 +458,8 @@  static void __init imx6q_clocks_init(struct device_node *ccm_node)
 	clk[IMX6QDL_CLK_PLL5_POST_DIV] = clk_register_divider_table(NULL, "pll5_post_div", "pll5_video", CLK_SET_RATE_PARENT, base + 0xa0, 19, 2, 0, post_div_table, &imx_ccm_lock);
 	clk[IMX6QDL_CLK_PLL5_VIDEO_DIV] = clk_register_divider_table(NULL, "pll5_video_div", "pll5_post_div", CLK_SET_RATE_PARENT, base + 0x170, 30, 2, 0, video_div_table, &imx_ccm_lock);
 
+	disable_anatop_clocks(base);
+
 	np = ccm_node;
 	base = of_iomap(np, 0);
 	WARN_ON(!base);
@@ -296,6 +487,16 @@  static void __init imx6q_clocks_init(struct device_node *ccm_node)
 	clk[IMX6QDL_CLK_GPU3D_SHADER_SEL] = imx_clk_mux("gpu3d_shader_sel", base + 0x18, 8,  2, gpu3d_shader_sels, ARRAY_SIZE(gpu3d_shader_sels));
 	clk[IMX6QDL_CLK_IPU1_SEL]         = imx_clk_mux("ipu1_sel",         base + 0x3c, 9,  2, ipu_sels,          ARRAY_SIZE(ipu_sels));
 	clk[IMX6QDL_CLK_IPU2_SEL]         = imx_clk_mux("ipu2_sel",         base + 0x3c, 14, 2, ipu_sels,          ARRAY_SIZE(ipu_sels));
+
+	/*
+	 * The LDB_DI0/1_SEL muxes are registered read-only due to a hardware
+	 * bug. Set the muxes to the requested values before registering the
+	 * ldb_di_sel clocks.
+	 */
+	if ((imx_get_soc_revision() != IMX_CHIP_REVISION_1_0) ||
+	    cpu_is_imx6dl())
+		init_ldb_clks(np, base);
+
 	clk[IMX6QDL_CLK_LDB_DI0_SEL]      = imx_clk_mux_ldb("ldb_di0_sel", base + 0x2c, 9,  3, ldb_di_sels,      ARRAY_SIZE(ldb_di_sels));
 	clk[IMX6QDL_CLK_LDB_DI1_SEL]      = imx_clk_mux_ldb("ldb_di1_sel", base + 0x2c, 12, 3, ldb_di_sels,      ARRAY_SIZE(ldb_di_sels));
 	clk[IMX6QDL_CLK_IPU1_DI0_PRE_SEL] = imx_clk_mux_flags("ipu1_di0_pre_sel", base + 0x34, 6,  3, ipu_di_pre_sels,   ARRAY_SIZE(ipu_di_pre_sels), CLK_SET_RATE_PARENT);