[v2] clk: samsung: Keep top BPLL mux on Exynos542x enabled
diff mbox series

Message ID 20200807133143.22748-1-m.szyprowski@samsung.com
State Not Applicable
Headers show
Series
  • [v2] clk: samsung: Keep top BPLL mux on Exynos542x enabled
Related show

Commit Message

Marek Szyprowski Aug. 7, 2020, 1:31 p.m. UTC
BPLL clock must not be disabled because it is needed for proper DRAM
operation. This is normally handled by respective memory devfreq driver,
but when that driver is not yet probed or its probe has been deferred the
clock might got disabled what causes board hang. Fix this by calling
clk_prepare_enable() directly from the clock provider driver.

Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
Reviewed-by: Lukasz Luba <lukasz.luba@arm.com>
Tested-by: Lukasz Luba <lukasz.luba@arm.com>
Acked-by: Krzysztof Kozlowski <krzk@kernel.org>
---
 drivers/clk/samsung/clk-exynos5420.c | 5 +++++
 1 file changed, 5 insertions(+)

Comments

Chanwoo Choi Aug. 10, 2020, 2:58 a.m. UTC | #1
Hi Marek,

On 8/7/20 10:31 PM, Marek Szyprowski wrote:
> BPLL clock must not be disabled because it is needed for proper DRAM
> operation. This is normally handled by respective memory devfreq driver,
> but when that driver is not yet probed or its probe has been deferred the
> clock might got disabled what causes board hang. Fix this by calling
> clk_prepare_enable() directly from the clock provider driver.
> 
> Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
> Reviewed-by: Lukasz Luba <lukasz.luba@arm.com>
> Tested-by: Lukasz Luba <lukasz.luba@arm.com>
> Acked-by: Krzysztof Kozlowski <krzk@kernel.org>
> ---
>  drivers/clk/samsung/clk-exynos5420.c | 5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/drivers/clk/samsung/clk-exynos5420.c b/drivers/clk/samsung/clk-exynos5420.c
> index fea33399a632..521cbbfc0987 100644
> --- a/drivers/clk/samsung/clk-exynos5420.c
> +++ b/drivers/clk/samsung/clk-exynos5420.c
> @@ -1655,6 +1655,11 @@ static void __init exynos5x_clk_init(struct device_node *np,
>  	 * main G3D clock enablement status.
>  	 */
>  	clk_prepare_enable(__clk_lookup("mout_sw_aclk_g3d"));
> +	/*
> +	 * Keep top BPLL mux enabled permanently to ensure that DRAM operates
> +	 * properly.
> +	 */
> +	clk_prepare_enable(__clk_lookup("mout_bpll"));
>  
>  	samsung_clk_of_add_provider(np, ctx);
>  }
> 

Thanks.

Acked-by: Chanwoo Choi <cw00.choi@samsung.com>
Sylwester Nawrocki Aug. 11, 2020, 11:31 a.m. UTC | #2
On 8/7/20 15:31, Marek Szyprowski wrote:
> BPLL clock must not be disabled because it is needed for proper DRAM
> operation. This is normally handled by respective memory devfreq driver,
> but when that driver is not yet probed or its probe has been deferred the
> clock might got disabled what causes board hang. Fix this by calling
> clk_prepare_enable() directly from the clock provider driver.
> 
> Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
> Reviewed-by: Lukasz Luba <lukasz.luba@arm.com>
> Tested-by: Lukasz Luba <lukasz.luba@arm.com>
> Acked-by: Krzysztof Kozlowski <krzk@kernel.org>

Should we add a "Fixes" tag so this commit gets backported down do the 
kernels where the DMC driver was introduced?

Fixes: 6e7674c3c6df ("memory: Add DMC driver for Exynos5422") ?

> ---
>   drivers/clk/samsung/clk-exynos5420.c | 5 +++++
>   1 file changed, 5 insertions(+)
> 
> diff --git a/drivers/clk/samsung/clk-exynos5420.c b/drivers/clk/samsung/clk-exynos5420.c
> index fea33399a632..521cbbfc0987 100644
> --- a/drivers/clk/samsung/clk-exynos5420.c
> +++ b/drivers/clk/samsung/clk-exynos5420.c
> @@ -1655,6 +1655,11 @@ static void __init exynos5x_clk_init(struct device_node *np,
>   	 * main G3D clock enablement status.
>   	 */
>   	clk_prepare_enable(__clk_lookup("mout_sw_aclk_g3d"));
> +	/*
> +	 * Keep top BPLL mux enabled permanently to ensure that DRAM operates
> +	 * properly.
> +	 */
> +	clk_prepare_enable(__clk_lookup("mout_bpll"));

I'm going to apply the patch and then these two as a follow up:

https://patchwork.kernel.org/patch/11709097
https://patchwork.kernel.org/patch/11709101
Stephen Boyd Aug. 19, 2020, 3:12 a.m. UTC | #3
Quoting Sylwester Nawrocki (2020-08-11 04:31:30)
> On 8/7/20 15:31, Marek Szyprowski wrote:
> > BPLL clock must not be disabled because it is needed for proper DRAM
> > operation. This is normally handled by respective memory devfreq driver,
> > but when that driver is not yet probed or its probe has been deferred the
> > clock might got disabled what causes board hang. Fix this by calling
> > clk_prepare_enable() directly from the clock provider driver.
> > 
> > Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
> > Reviewed-by: Lukasz Luba <lukasz.luba@arm.com>
> > Tested-by: Lukasz Luba <lukasz.luba@arm.com>
> > Acked-by: Krzysztof Kozlowski <krzk@kernel.org>
> 
> Should we add a "Fixes" tag so this commit gets backported down do the 
> kernels where the DMC driver was introduced?
> 
> Fixes: 6e7674c3c6df ("memory: Add DMC driver for Exynos5422") ?

I've recently discovered that stable trees aren't checking for Fixes
tags. So we have to put both a Fixes tag and a Cc stable on the patch
to make sure it gets applied to stable trees. Otherwise it's up to the
robot to figure out that a Fixes tag means maybe this is important.
Stephen Boyd Aug. 19, 2020, 3:14 a.m. UTC | #4
Quoting Marek Szyprowski (2020-08-07 06:31:43)
> BPLL clock must not be disabled because it is needed for proper DRAM
> operation. This is normally handled by respective memory devfreq driver,
> but when that driver is not yet probed or its probe has been deferred the
> clock might got disabled what causes board hang. Fix this by calling
> clk_prepare_enable() directly from the clock provider driver.
> 
> Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
> Reviewed-by: Lukasz Luba <lukasz.luba@arm.com>
> Tested-by: Lukasz Luba <lukasz.luba@arm.com>
> Acked-by: Krzysztof Kozlowski <krzk@kernel.org>
> ---

Can I pick this up for clk-fixes?
Sylwester Nawrocki Aug. 23, 2020, 10:12 a.m. UTC | #5
On 8/19/20 05:14, Stephen Boyd wrote:
> Quoting Marek Szyprowski (2020-08-07 06:31:43)
>> BPLL clock must not be disabled because it is needed for proper DRAM
>> operation. This is normally handled by respective memory devfreq driver,
>> but when that driver is not yet probed or its probe has been deferred the
>> clock might got disabled what causes board hang. Fix this by calling
>> clk_prepare_enable() directly from the clock provider driver.
>>
>> Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
>> Reviewed-by: Lukasz Luba <lukasz.luba@arm.com>
>> Tested-by: Lukasz Luba <lukasz.luba@arm.com>
>> Acked-by: Krzysztof Kozlowski <krzk@kernel.org>
>> ---
> 
> Can I pick this up for clk-fixes?

Sure, thanks for taking care of this.
Sylwester Nawrocki Aug. 24, 2020, 10:28 a.m. UTC | #6
On 8/23/20 12:12, Sylwester Nawrocki wrote:
> On 8/19/20 05:14, Stephen Boyd wrote:
>> Quoting Marek Szyprowski (2020-08-07 06:31:43)
>>> BPLL clock must not be disabled because it is needed for proper DRAM
>>> operation. This is normally handled by respective memory devfreq driver,
>>> but when that driver is not yet probed or its probe has been deferred 
>>> the
>>> clock might got disabled what causes board hang. Fix this by calling
>>> clk_prepare_enable() directly from the clock provider driver.
>>>
>>> Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
>>> Reviewed-by: Lukasz Luba <lukasz.luba@arm.com>
>>> Tested-by: Lukasz Luba <lukasz.luba@arm.com>
>>> Acked-by: Krzysztof Kozlowski <krzk@kernel.org>
>>> ---
>>
>> Can I pick this up for clk-fixes?
> 
> Sure, thanks for taking care of this.

OTOH, I planned to queue that patch for next merged window, together 
with a patch that depends on that one, since the fix is not for an issue
introduced in the last merge window.
I guess it's better to avoid pulling (part of) the clk-fixes branch to
the clk/samsung tree for next merge window?
Krzysztof Kozlowski Aug. 24, 2020, 10:31 a.m. UTC | #7
On Mon, Aug 24, 2020 at 12:28:51PM +0200, Sylwester Nawrocki wrote:
> On 8/23/20 12:12, Sylwester Nawrocki wrote:
> > On 8/19/20 05:14, Stephen Boyd wrote:
> > > Quoting Marek Szyprowski (2020-08-07 06:31:43)
> > > > BPLL clock must not be disabled because it is needed for proper DRAM
> > > > operation. This is normally handled by respective memory devfreq driver,
> > > > but when that driver is not yet probed or its probe has been
> > > > deferred the
> > > > clock might got disabled what causes board hang. Fix this by calling
> > > > clk_prepare_enable() directly from the clock provider driver.
> > > > 
> > > > Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
> > > > Reviewed-by: Lukasz Luba <lukasz.luba@arm.com>
> > > > Tested-by: Lukasz Luba <lukasz.luba@arm.com>
> > > > Acked-by: Krzysztof Kozlowski <krzk@kernel.org>
> > > > ---
> > > 
> > > Can I pick this up for clk-fixes?
> > 
> > Sure, thanks for taking care of this.
> 
> OTOH, I planned to queue that patch for next merged window, together with a
> patch that depends on that one, since the fix is not for an issue
> introduced in the last merge window.
> I guess it's better to avoid pulling (part of) the clk-fixes branch to
> the clk/samsung tree for next merge window?

All current multi_v7 and some of exynos defconfig boots fail on Odroid
XU3-family, starting from v5.9-rc1. On kernelci and my boot systems. If
I understand correctly, this is a fix for this issue, so it should go as
fast as possible to v5.9 cycle.

Otherwise we cannot test anything. The current v5.9 RC is then simply broken.

Best regards,
Krzysztof
Sylwester Nawrocki Sept. 2, 2020, 9:24 a.m. UTC | #8
On 24.08.2020 12:31, Krzysztof Kozlowski wrote:
> On Mon, Aug 24, 2020 at 12:28:51PM +0200, Sylwester Nawrocki wrote:
>> On 8/23/20 12:12, Sylwester Nawrocki wrote:
>>> On 8/19/20 05:14, Stephen Boyd wrote:
>>>> Quoting Marek Szyprowski (2020-08-07 06:31:43)
>>>>> BPLL clock must not be disabled because it is needed for proper DRAM
>>>>> operation. This is normally handled by respective memory devfreq driver,
>>>>> but when that driver is not yet probed or its probe has been
>>>>> deferred the
>>>>> clock might got disabled what causes board hang. Fix this by calling
>>>>> clk_prepare_enable() directly from the clock provider driver.
>>>>>
>>>>> Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
>>>>> Reviewed-by: Lukasz Luba <lukasz.luba@arm.com>
>>>>> Tested-by: Lukasz Luba <lukasz.luba@arm.com>
>>>>> Acked-by: Krzysztof Kozlowski <krzk@kernel.org>
>>>>> ---
>>>>
>>>> Can I pick this up for clk-fixes?
>>>
>>> Sure, thanks for taking care of this.
>>
>> OTOH, I planned to queue that patch for next merged window, together 
>> with a patch that depends on that one, since the fix is not for an issue
>> introduced in the last merge window.
>> I guess it's better to avoid pulling (part of) the clk-fixes branch to
>> the clk/samsung tree for next merge window?
> 
> All current multi_v7 and some of exynos defconfig boots fail on Odroid
> XU3-family, starting from v5.9-rc1. On kernelci and my boot systems. If
> I understand correctly, this is a fix for this issue, so it should go as
> fast as possible to v5.9 cycle.
> 
> Otherwise we cannot test anything. The current v5.9 RC is then simply
> broken.

Right, we need that patch in v5.9. Stephen, can you please apply
the patch to your clk-fixes?

--
Thanks, 
Sylwester
Sylwester Nawrocki Sept. 15, 2020, 12:43 p.m. UTC | #9
On 02.09.2020 11:24, Sylwester Nawrocki wrote:
> On 24.08.2020 12:31, Krzysztof Kozlowski wrote:
>> On Mon, Aug 24, 2020 at 12:28:51PM +0200, Sylwester Nawrocki wrote:
>>> On 8/23/20 12:12, Sylwester Nawrocki wrote:
>>>> On 8/19/20 05:14, Stephen Boyd wrote:
>>>>> Quoting Marek Szyprowski (2020-08-07 06:31:43)
>>>>>> BPLL clock must not be disabled because it is needed for proper DRAM
>>>>>> operation. This is normally handled by respective memory devfreq driver,
>>>>>> but when that driver is not yet probed or its probe has been
>>>>>> deferred the
>>>>>> clock might got disabled what causes board hang. Fix this by calling
>>>>>> clk_prepare_enable() directly from the clock provider driver.
>>>>>>
>>>>>> Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
>>>>>> Reviewed-by: Lukasz Luba <lukasz.luba@arm.com>
>>>>>> Tested-by: Lukasz Luba <lukasz.luba@arm.com>
>>>>>> Acked-by: Krzysztof Kozlowski <krzk@kernel.org>
>>>>>> ---
>>>>>
>>>>> Can I pick this up for clk-fixes?
>>>>
>>>> Sure, thanks for taking care of this.
>>>
>>> OTOH, I planned to queue that patch for next merged window, together 
>>> with a patch that depends on that one, since the fix is not for an issue
>>> introduced in the last merge window.
>>> I guess it's better to avoid pulling (part of) the clk-fixes branch to
>>> the clk/samsung tree for next merge window?
>>
>> All current multi_v7 and some of exynos defconfig boots fail on Odroid
>> XU3-family, starting from v5.9-rc1. On kernelci and my boot systems. If
>> I understand correctly, this is a fix for this issue, so it should go as
>> fast as possible to v5.9 cycle.
>>
>> Otherwise we cannot test anything. The current v5.9 RC is then simply
>> broken.
> 
> Right, we need that patch in v5.9. Stephen, can you please apply
> the patch to your clk-fixes?

So I applied the patch to my tree and sent you a pull request
instead... :) I thought it will handling subsequent patches
that depend on that one more straightforward.
Stephen Boyd Sept. 15, 2020, 4:17 p.m. UTC | #10
Quoting Sylwester Nawrocki (2020-09-15 05:43:07)
> On 02.09.2020 11:24, Sylwester Nawrocki wrote:
> > On 24.08.2020 12:31, Krzysztof Kozlowski wrote:
> >>> the clk/samsung tree for next merge window?
> >>
> >> All current multi_v7 and some of exynos defconfig boots fail on Odroid
> >> XU3-family, starting from v5.9-rc1. On kernelci and my boot systems. If
> >> I understand correctly, this is a fix for this issue, so it should go as
> >> fast as possible to v5.9 cycle.
> >>
> >> Otherwise we cannot test anything. The current v5.9 RC is then simply
> >> broken.
> > 
> > Right, we need that patch in v5.9. Stephen, can you please apply
> > the patch to your clk-fixes?
> 
> So I applied the patch to my tree and sent you a pull request
> instead... :) I thought it will handling subsequent patches
> that depend on that one more straightforward.
> 

Great!

Patch
diff mbox series

diff --git a/drivers/clk/samsung/clk-exynos5420.c b/drivers/clk/samsung/clk-exynos5420.c
index fea33399a632..521cbbfc0987 100644
--- a/drivers/clk/samsung/clk-exynos5420.c
+++ b/drivers/clk/samsung/clk-exynos5420.c
@@ -1655,6 +1655,11 @@  static void __init exynos5x_clk_init(struct device_node *np,
 	 * main G3D clock enablement status.
 	 */
 	clk_prepare_enable(__clk_lookup("mout_sw_aclk_g3d"));
+	/*
+	 * Keep top BPLL mux enabled permanently to ensure that DRAM operates
+	 * properly.
+	 */
+	clk_prepare_enable(__clk_lookup("mout_bpll"));
 
 	samsung_clk_of_add_provider(np, ctx);
 }