Skip to contentSkip to conversation

mmc: sdhci-xenon: fix 1.8v regulator stabilization

PATCH In review Marcin Wojtas 4 participants 11 messages Dec 12, 2020 – Mar 18, 2022

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 mmc: sdhci-xenon: fix 1.8v regulator stabilization

From: Alex Leibovich <[email protected]>

Automatic Clock Gating is a feature used for the power consumption optimisation. It turned out that during early init phase it may prevent the stable voltage switch to 1.8V - due to that on some platfroms an endless printout in dmesg can be observed:
"mmc1: 1.8V regulator output did not became stable" Fix the problem by disabling the ACG at very beginning of the sdhci_init and let that be enabled later.

Fixes 3a3748dba881 ("mmc: sdhci-xenon: Add Marvell Xenon SDHC core functionality")
---
 drivers/mmc/host/sdhci-xenon.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
drivers/mmc/host/sdhci-xenon.c +6 -1
index c67611fdaa8a..4b05f6fdefb4 100644
@@ -168,7 +168,12 @@ static void xenon_reset_exit(struct sdhci_host *host,
 	/* Disable tuning request and auto-retuning again */
 	xenon_retune_setup(host);
 
-	xenon_set_acg(host, true);
+	/*
+	 * The ACG should be turned off at the early init time, in order
+	 * to solve a possile issues with the 1.8V regulator stabilization.
+	 * The feature is enabled in later stage.
+	 */
+	xenon_set_acg(host, false);
 
 	xenon_set_sdclk_off_idle(host, sdhc_id, false);
 
Signature · 1 line
2.29.0
maintainer Acked-by on v1 1/1
#1 lore raw
Quoted 6 lines · Marcin Wojtas

From: Alex Leibovich <[email protected]>

Automatic Clock Gating is a feature used for the power consumption optimisation. It turned out that during early init phase it may prevent the stable voltage switch to 1.8V - due to that on some platfroms an endless

platfroms -> platforms

Quoted 9 lines

printout in dmesg can be observed:
"mmc1: 1.8V regulator output did not became stable" Fix the problem by disabling the ACG at very beginning of the sdhci_init and let that be enabled later.

Fixes 3a3748dba881 ("mmc: sdhci-xenon: Add Marvell Xenon SDHC core functionality")
Quoted 9 lines
---
 drivers/mmc/host/sdhci-xenon.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
drivers/mmc/host/sdhci-xenon.c +0 -0
index c67611fdaa8a..4b05f6fdefb4 100644

drivers/mmc/host/sdhci-xenon.c static void xenon_reset_exit(struct sdhci_host *host, · line 173

Quoted 7 lines
drivers/mmc/host/sdhci-xenon.c +3 -1
@@ -168,7 +168,12 @@ static void xenon_reset_exit(struct sdhci_host *host,
 	/* Disable tuning request and auto-retuning again */
 	xenon_retune_setup(host);
 
-	xenon_set_acg(host, true);
+	/*
+	 * The ACG should be turned off at the early init time, in order
+	 * to solve a possile issues with the 1.8V regulator stabilization.

a possile -> possible

Quoted 7 lines
drivers/mmc/host/sdhci-xenon.c +3 -0
+	 * The feature is enabled in later stage.
+	 */
+	xenon_set_acg(host, false);
 
 	xenon_set_sdclk_off_idle(host, sdhc_id, false);
 
#2 lore raw

wt., 15 gru 2020 o 09:04 Adrian Hunter <[email protected]> napisał(a):

Quoted 16 lines · Marcin Wojtas
Quoted 6 lines

From: Alex Leibovich <[email protected]>

Automatic Clock Gating is a feature used for the power consumption optimisation. It turned out that during early init phase it may prevent the stable voltage switch to 1.8V - due to that on some platfroms an endless

platfroms -> platforms

Quoted 5 lines

printout in dmesg can be observed:
"mmc1: 1.8V regulator output did not became stable" Fix the problem by disabling the ACG at very beginning of the sdhci_init and let that be enabled later.

Quoted 7 lines
Quoted 4 lines
Fixes 3a3748dba881 ("mmc: sdhci-xenon: Add Marvell Xenon SDHC core functionality")

Thank you, I'll repost right away with the typos correction.

Best regards,
Marcin

Quoted 27 lines
Quoted 9 lines
---
 drivers/mmc/host/sdhci-xenon.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
drivers/mmc/host/sdhci-xenon.c +0 -0
index c67611fdaa8a..4b05f6fdefb4 100644

drivers/mmc/host/sdhci-xenon.c static void xenon_reset_exit(struct sdhci_host *host, · line 173

Quoted 7 lines
drivers/mmc/host/sdhci-xenon.c +3 -1
@@ -168,7 +168,12 @@ static void xenon_reset_exit(struct sdhci_host *host,
      /* Disable tuning request and auto-retuning again */
      xenon_retune_setup(host);
 
-     xenon_set_acg(host, true);
+     /*
+      * The ACG should be turned off at the early init time, in order
+      * to solve a possile issues with the 1.8V regulator stabilization.

a possile -> possible

Quoted 7 lines
drivers/mmc/host/sdhci-xenon.c +3 -0
+      * The feature is enabled in later stage.
+      */
+     xenon_set_acg(host, false);
 
      xenon_set_sdclk_off_idle(host, sdhc_id, false);
 
#3 lore raw
In reply to Marcin Wojtas
Quoted 12 lines · Marcin Wojtas

From: Alex Leibovich <[email protected]>

Automatic Clock Gating is a feature used for the power consumption optimisation. It turned out that during early init phase it may prevent the stable voltage switch to 1.8V - due to that on some platfroms an endless printout in dmesg can be observed:
"mmc1: 1.8V regulator output did not became stable" Fix the problem by disabling the ACG at very beginning of the sdhci_init and let that be enabled later.

Quoted 4 lines
Fixes 3a3748dba881 ("mmc: sdhci-xenon: Add Marvell Xenon SDHC core functionality")

Applied for fixes (by fixing the typos), thanks!

Kind regards
Uffe

Quoted 25 lines
---
 drivers/mmc/host/sdhci-xenon.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
drivers/mmc/host/sdhci-xenon.c +6 -2
index c67611fdaa8a..4b05f6fdefb4 100644
@@ -168,7 +168,12 @@ static void xenon_reset_exit(struct sdhci_host *host,
        /* Disable tuning request and auto-retuning again */
        xenon_retune_setup(host);
 
-       xenon_set_acg(host, true);
+       /*
+        * The ACG should be turned off at the early init time, in order
+        * to solve a possile issues with the 1.8V regulator stabilization.
+        * The feature is enabled in later stage.
+        */
+       xenon_set_acg(host, false);
 
        xenon_set_sdclk_off_idle(host, sdhc_id, false);
 
--

2.29.0

#4 lore raw
Quoted 13 lines · On Monday 11 January 2021 19:06:24 Ulf Hansson wrote:
Quoted 12 lines · Marcin Wojtas

From: Alex Leibovich <[email protected]>

Automatic Clock Gating is a feature used for the power consumption optimisation. It turned out that during early init phase it may prevent the stable voltage switch to 1.8V - due to that on some platfroms an endless printout in dmesg can be observed:
"mmc1: 1.8V regulator output did not became stable" Fix the problem by disabling the ACG at very beginning of the sdhci_init and let that be enabled later.

Quoted 6 lines
Quoted 4 lines
Fixes 3a3748dba881 ("mmc: sdhci-xenon: Add Marvell Xenon SDHC core functionality")

Applied for fixes (by fixing the typos), thanks!

Hello!

Is not this patch address same issue which was fixed by patch which was merged earlier?

bb32e1987bc5 ("mmc: sdhci-xenon: fix annoying 1.8V regulator warning")
https://lore.kernel.org/linux-mmc/CAPDyKFqAsvgAjfL-c9ukFNWeGJmufQosR2Eg9SKjXMVpNitdkA@mail.gmail.com/

Quoted 29 lines

Kind regards
Uffe

Quoted 25 lines
---
 drivers/mmc/host/sdhci-xenon.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
drivers/mmc/host/sdhci-xenon.c +6 -2
index c67611fdaa8a..4b05f6fdefb4 100644
@@ -168,7 +168,12 @@ static void xenon_reset_exit(struct sdhci_host *host,
        /* Disable tuning request and auto-retuning again */
        xenon_retune_setup(host);
 
-       xenon_set_acg(host, true);
+       /*
+        * The ACG should be turned off at the early init time, in order
+        * to solve a possile issues with the 1.8V regulator stabilization.
+        * The feature is enabled in later stage.
+        */
+       xenon_set_acg(host, false);
 
        xenon_set_sdclk_off_idle(host, sdhc_id, false);
 
--

2.29.0

#5 lore raw

Hi Pali,

pon., 14 mar 2022 o 16:40 Pali Rohár <[email protected]> napisał(a):

Quoted 22 lines · On Monday 11 January 2021 19:06:24 Ulf Hansson wrote:
Quoted 13 lines · Marcin Wojtas
Quoted 12 lines

From: Alex Leibovich <[email protected]>

Automatic Clock Gating is a feature used for the power consumption optimisation. It turned out that during early init phase it may prevent the stable voltage switch to 1.8V - due to that on some platfroms an endless printout in dmesg can be observed:
"mmc1: 1.8V regulator output did not became stable" Fix the problem by disabling the ACG at very beginning of the sdhci_init and let that be enabled later.

Quoted 6 lines
Quoted 4 lines
Fixes 3a3748dba881 ("mmc: sdhci-xenon: Add Marvell Xenon SDHC core functionality")

Applied for fixes (by fixing the typos), thanks!

Quoted 8 lines

Hello!

Is not this patch address same issue which was fixed by patch which was merged earlier?

bb32e1987bc5 ("mmc: sdhci-xenon: fix annoying 1.8V regulator warning")
https://lore.kernel.org/linux-mmc/CAPDyKFqAsvgAjfL-c9ukFNWeGJmufQosR2Eg9SKjXMVpNitdkA@mail.gmail.com/

This indeed look similar. This fix was originally developed for CN913x platform without the mentioned patch (I'm wondering if it would also suffice to fix A3k board's problem). Anyway, I don't think we have an issue here, as everything seems to work fine on top of mainline Linux with both changes.

Best regards,
Marcin

Quoted 29 lines
Quoted 29 lines

Kind regards
Uffe

Quoted 25 lines
---
 drivers/mmc/host/sdhci-xenon.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
drivers/mmc/host/sdhci-xenon.c +6 -2
index c67611fdaa8a..4b05f6fdefb4 100644
@@ -168,7 +168,12 @@ static void xenon_reset_exit(struct sdhci_host *host,
        /* Disable tuning request and auto-retuning again */
        xenon_retune_setup(host);
 
-       xenon_set_acg(host, true);
+       /*
+        * The ACG should be turned off at the early init time, in order
+        * to solve a possile issues with the 1.8V regulator stabilization.
+        * The feature is enabled in later stage.
+        */
+       xenon_set_acg(host, false);
 
        xenon_set_sdclk_off_idle(host, sdhc_id, false);
 
--

2.29.0

#6 lore raw

Hello!

Quoted 35 lines · On Monday 14 March 2022 16:51:25 Marcin Wojtas wrote:

Hi Pali,

pon., 14 mar 2022 o 16:40 Pali Rohár <[email protected]> napisał(a):

Quoted 22 lines · On Monday 11 January 2021 19:06:24 Ulf Hansson wrote:
Quoted 13 lines · Marcin Wojtas
Quoted 12 lines

From: Alex Leibovich <[email protected]>

Automatic Clock Gating is a feature used for the power consumption optimisation. It turned out that during early init phase it may prevent the stable voltage switch to 1.8V - due to that on some platfroms an endless printout in dmesg can be observed:
"mmc1: 1.8V regulator output did not became stable" Fix the problem by disabling the ACG at very beginning of the sdhci_init and let that be enabled later.

Quoted 6 lines
Quoted 4 lines
Fixes 3a3748dba881 ("mmc: sdhci-xenon: Add Marvell Xenon SDHC core functionality")

Applied for fixes (by fixing the typos), thanks!

Quoted 8 lines

Hello!

Is not this patch address same issue which was fixed by patch which was merged earlier?

bb32e1987bc5 ("mmc: sdhci-xenon: fix annoying 1.8V regulator warning")
https://lore.kernel.org/linux-mmc/CAPDyKFqAsvgAjfL-c9ukFNWeGJmufQosR2Eg9SKjXMVpNitdkA@mail.gmail.com/

Quoted 5 lines

This indeed look similar. This fix was originally developed for CN913x platform without the mentioned patch (I'm wondering if it would also suffice to fix A3k board's problem). Anyway, I don't think we have an issue here, as everything seems to work fine on top of mainline Linux with both changes.

Yea, there should be no issue. Just question is if we need _both_ fixes.

I could probably try to revert bb32e1987bc5 and check what happens on A3k board.

Quoted 32 lines

Best regards,
Marcin

Quoted 29 lines
Quoted 29 lines

Kind regards
Uffe

Quoted 25 lines
---
 drivers/mmc/host/sdhci-xenon.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
drivers/mmc/host/sdhci-xenon.c +6 -2
index c67611fdaa8a..4b05f6fdefb4 100644
@@ -168,7 +168,12 @@ static void xenon_reset_exit(struct sdhci_host *host,
        /* Disable tuning request and auto-retuning again */
        xenon_retune_setup(host);
 
-       xenon_set_acg(host, true);
+       /*
+        * The ACG should be turned off at the early init time, in order
+        * to solve a possile issues with the 1.8V regulator stabilization.
+        * The feature is enabled in later stage.
+        */
+       xenon_set_acg(host, false);
 
        xenon_set_sdclk_off_idle(host, sdhc_id, false);
 
--

2.29.0

#7 lore raw

Hi Pali,

śr., 16 mar 2022 o 00:03 Pali Rohár <[email protected]> napisał(a):

Quoted 45 lines

Hello!

Quoted 35 lines · On Monday 14 March 2022 16:51:25 Marcin Wojtas wrote:

Hi Pali,

pon., 14 mar 2022 o 16:40 Pali Rohár <[email protected]> napisał(a):

Quoted 22 lines · On Monday 11 January 2021 19:06:24 Ulf Hansson wrote:
Quoted 13 lines · Marcin Wojtas
Quoted 12 lines

From: Alex Leibovich <[email protected]>

Automatic Clock Gating is a feature used for the power consumption optimisation. It turned out that during early init phase it may prevent the stable voltage switch to 1.8V - due to that on some platfroms an endless printout in dmesg can be observed:
"mmc1: 1.8V regulator output did not became stable" Fix the problem by disabling the ACG at very beginning of the sdhci_init and let that be enabled later.

Quoted 6 lines
Quoted 4 lines
Fixes 3a3748dba881 ("mmc: sdhci-xenon: Add Marvell Xenon SDHC core functionality")

Applied for fixes (by fixing the typos), thanks!

Quoted 8 lines

Hello!

Is not this patch address same issue which was fixed by patch which was merged earlier?

bb32e1987bc5 ("mmc: sdhci-xenon: fix annoying 1.8V regulator warning")
https://lore.kernel.org/linux-mmc/CAPDyKFqAsvgAjfL-c9ukFNWeGJmufQosR2Eg9SKjXMVpNitdkA@mail.gmail.com/

Quoted 5 lines

This indeed look similar. This fix was originally developed for CN913x platform without the mentioned patch (I'm wondering if it would also suffice to fix A3k board's problem). Anyway, I don't think we have an issue here, as everything seems to work fine on top of mainline Linux with both changes.

Quoted 5 lines

Yea, there should be no issue. Just question is if we need _both_ fixes.

I could probably try to revert bb32e1987bc5 and check what happens on A3k board.

Yes, that would be interesting. Please let me know whenever you find time to check.

Best regards,
Marcin

Quoted 27 lines
Quoted 27 lines
Quoted 27 lines
Quoted 27 lines
Quoted 27 lines

---
 drivers/mmc/host/sdhci-xenon.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
drivers/mmc/host/sdhci-xenon.c +6 -2
index c67611fdaa8a..4b05f6fdefb4 100644
@@ -168,7 +168,12 @@ static void xenon_reset_exit(struct sdhci_host *host,
        /* Disable tuning request and auto-retuning again */
        xenon_retune_setup(host);
 
-       xenon_set_acg(host, true);
+       /*
+        * The ACG should be turned off at the early init time, in order
+        * to solve a possile issues with the 1.8V regulator stabilization.
+        * The feature is enabled in later stage.
+        */
+       xenon_set_acg(host, false);
 
        xenon_set_sdclk_off_idle(host, sdhc_id, false);
 
--

2.29.0

#8 lore raw
Quoted 48 lines · On Wednesday 16 March 2022 02:03:35 Marcin Wojtas wrote:

Hi Pali,

śr., 16 mar 2022 o 00:03 Pali Rohár <[email protected]> napisał(a):

Quoted 45 lines

Hello!

Quoted 40 lines · On Monday 14 March 2022 16:51:25 Marcin Wojtas wrote:

Hi Pali,

pon., 14 mar 2022 o 16:40 Pali Rohár <[email protected]> napisał(a):

Quoted 22 lines · On Monday 11 January 2021 19:06:24 Ulf Hansson wrote:
Quoted 13 lines · Marcin Wojtas
Quoted 12 lines

From: Alex Leibovich <[email protected]>

Automatic Clock Gating is a feature used for the power consumption optimisation. It turned out that during early init phase it may prevent the stable voltage switch to 1.8V - due to that on some platfroms an endless printout in dmesg can be observed:
"mmc1: 1.8V regulator output did not became stable" Fix the problem by disabling the ACG at very beginning of the sdhci_init and let that be enabled later.

Quoted 6 lines
Quoted 4 lines
Fixes 3a3748dba881 ("mmc: sdhci-xenon: Add Marvell Xenon SDHC core functionality")

Applied for fixes (by fixing the typos), thanks!

Quoted 8 lines

Hello!

Is not this patch address same issue which was fixed by patch which was merged earlier?

bb32e1987bc5 ("mmc: sdhci-xenon: fix annoying 1.8V regulator warning")
https://lore.kernel.org/linux-mmc/CAPDyKFqAsvgAjfL-c9ukFNWeGJmufQosR2Eg9SKjXMVpNitdkA@mail.gmail.com/

This indeed look similar. This fix was originally developed for CN913x platform without the mentioned patch (I'm wondering if it would also suffice to fix A3k board's problem). Anyway, I don't think we have an issue here, as everything seems to work fine on top of mainline Linux with both changes.

Quoted 8 lines
Quoted 5 lines

Yea, there should be no issue. Just question is if we need _both_ fixes.

I could probably try to revert bb32e1987bc5 and check what happens on A3k board.

Yes, that would be interesting. Please let me know whenever you find time to check.

Hello! Now I tested kernel with reverted commit bb32e1987bc5 ("mmc:
sdhci-xenon: fix annoying 1.8V regulator warning") and issue is still fixed. I reverted also bb32e1987bc5 ("mmc: sdhci-xenon: fix annoying 1.8V regulator warning") commit and then issue appeared again.

So any of this commit is fixing that issue on Armada 3720.

Should we revert one of them?

Quoted 30 lines

Best regards,
Marcin

Quoted 27 lines
Quoted 27 lines
Quoted 27 lines
Quoted 27 lines
Quoted 27 lines

---
 drivers/mmc/host/sdhci-xenon.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
drivers/mmc/host/sdhci-xenon.c +6 -2
index c67611fdaa8a..4b05f6fdefb4 100644
@@ -168,7 +168,12 @@ static void xenon_reset_exit(struct sdhci_host *host,
        /* Disable tuning request and auto-retuning again */
        xenon_retune_setup(host);
 
-       xenon_set_acg(host, true);
+       /*
+        * The ACG should be turned off at the early init time, in order
+        * to solve a possile issues with the 1.8V regulator stabilization.
+        * The feature is enabled in later stage.
+        */
+       xenon_set_acg(host, false);
 
        xenon_set_sdclk_off_idle(host, sdhc_id, false);
 
--

2.29.0

#9 lore raw
Quoted 55 lines · On Friday 18 March 2022 14:01:00 Pali Rohár wrote:
Quoted 54 lines · On Wednesday 16 March 2022 02:03:35 Marcin Wojtas wrote:

Hi Pali,

śr., 16 mar 2022 o 00:03 Pali Rohár <[email protected]> napisał(a):

Quoted 50 lines

Hello!

Quoted 35 lines · On Monday 14 March 2022 16:51:25 Marcin Wojtas wrote:

Hi Pali,

pon., 14 mar 2022 o 16:40 Pali Rohár <[email protected]> napisał(a):

Quoted 22 lines · On Monday 11 January 2021 19:06:24 Ulf Hansson wrote:
Quoted 13 lines · Marcin Wojtas
Quoted 12 lines

From: Alex Leibovich <[email protected]>

Automatic Clock Gating is a feature used for the power consumption optimisation. It turned out that during early init phase it may prevent the stable voltage switch to 1.8V - due to that on some platfroms an endless printout in dmesg can be observed:
"mmc1: 1.8V regulator output did not became stable" Fix the problem by disabling the ACG at very beginning of the sdhci_init and let that be enabled later.

Quoted 6 lines
Quoted 4 lines
Fixes 3a3748dba881 ("mmc: sdhci-xenon: Add Marvell Xenon SDHC core functionality")

Applied for fixes (by fixing the typos), thanks!

Quoted 8 lines

Hello!

Is not this patch address same issue which was fixed by patch which was merged earlier?

bb32e1987bc5 ("mmc: sdhci-xenon: fix annoying 1.8V regulator warning")
https://lore.kernel.org/linux-mmc/CAPDyKFqAsvgAjfL-c9ukFNWeGJmufQosR2Eg9SKjXMVpNitdkA@mail.gmail.com/

Quoted 5 lines

This indeed look similar. This fix was originally developed for CN913x platform without the mentioned patch (I'm wondering if it would also suffice to fix A3k board's problem). Anyway, I don't think we have an issue here, as everything seems to work fine on top of mainline Linux with both changes.

Yea, there should be no issue. Just question is if we need _both_ fixes.

I could probably try to revert bb32e1987bc5 and check what happens on A3k board.

Quoted 7 lines
Quoted 2 lines

Yes, that would be interesting. Please let me know whenever you find time to check.

Hello! Now I tested kernel with reverted commit bb32e1987bc5 ("mmc:
sdhci-xenon: fix annoying 1.8V regulator warning") and issue is still fixed. I reverted also bb32e1987bc5 ("mmc: sdhci-xenon: fix annoying 1.8V regulator warning") commit and then issue appeared again.

I mean that I reverted also 1a3ed0dc3594 ("mmc: sdhci-xenon: fix 1.8v regulator stabilization") commit and then issue appeared again.

Quoted 34 lines

So any of this commit is fixing that issue on Armada 3720.

Should we revert one of them?

Quoted 30 lines

Best regards,
Marcin

Quoted 27 lines
Quoted 27 lines
Quoted 27 lines
Quoted 27 lines
Quoted 27 lines

---
 drivers/mmc/host/sdhci-xenon.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
drivers/mmc/host/sdhci-xenon.c +6 -2
index c67611fdaa8a..4b05f6fdefb4 100644
@@ -168,7 +168,12 @@ static void xenon_reset_exit(struct sdhci_host *host,
        /* Disable tuning request and auto-retuning again */
        xenon_retune_setup(host);
 
-       xenon_set_acg(host, true);
+       /*
+        * The ACG should be turned off at the early init time, in order
+        * to solve a possile issues with the 1.8V regulator stabilization.
+        * The feature is enabled in later stage.
+        */
+       xenon_set_acg(host, false);
 
        xenon_set_sdclk_off_idle(host, sdhc_id, false);
 
--

2.29.0

#10 lore raw

Hi Pali,

Thanks for testing!

pt., 18 mar 2022 o 14:06 Pali Rohár <[email protected]> napisał(a):

Quoted 60 lines · On Friday 18 March 2022 14:01:00 Pali Rohár wrote:
Quoted 58 lines · On Wednesday 16 March 2022 02:03:35 Marcin Wojtas wrote:
Quoted 56 lines

Hi Pali,

śr., 16 mar 2022 o 00:03 Pali Rohár <[email protected]> napisał(a):

Quoted 45 lines

Hello!

Quoted 35 lines · On Monday 14 March 2022 16:51:25 Marcin Wojtas wrote:

Hi Pali,

pon., 14 mar 2022 o 16:40 Pali Rohár <[email protected]> napisał(a):

Quoted 22 lines · On Monday 11 January 2021 19:06:24 Ulf Hansson wrote:
Quoted 13 lines · Marcin Wojtas
Quoted 12 lines

From: Alex Leibovich <[email protected]>

Automatic Clock Gating is a feature used for the power consumption optimisation. It turned out that during early init phase it may prevent the stable voltage switch to 1.8V - due to that on some platfroms an endless printout in dmesg can be observed:
"mmc1: 1.8V regulator output did not became stable" Fix the problem by disabling the ACG at very beginning of the sdhci_init and let that be enabled later.

Quoted 6 lines
Quoted 4 lines
Fixes 3a3748dba881 ("mmc: sdhci-xenon: Add Marvell Xenon SDHC core functionality")

Applied for fixes (by fixing the typos), thanks!

Quoted 8 lines

Hello!

Is not this patch address same issue which was fixed by patch which was merged earlier?

bb32e1987bc5 ("mmc: sdhci-xenon: fix annoying 1.8V regulator warning")
https://lore.kernel.org/linux-mmc/CAPDyKFqAsvgAjfL-c9ukFNWeGJmufQosR2Eg9SKjXMVpNitdkA@mail.gmail.com/

Quoted 5 lines

This indeed look similar. This fix was originally developed for CN913x platform without the mentioned patch (I'm wondering if it would also suffice to fix A3k board's problem). Anyway, I don't think we have an issue here, as everything seems to work fine on top of mainline Linux with both changes.

Quoted 5 lines

Yea, there should be no issue. Just question is if we need _both_ fixes.

I could probably try to revert bb32e1987bc5 and check what happens on A3k board.

Yes, that would be interesting. Please let me know whenever you find time to check.

Quoted 10 lines
Quoted 4 lines

Hello! Now I tested kernel with reverted commit bb32e1987bc5 ("mmc:
sdhci-xenon: fix annoying 1.8V regulator warning") and issue is still fixed. I reverted also bb32e1987bc5 ("mmc: sdhci-xenon: fix annoying 1.8V regulator warning") commit and then issue appeared again.

I mean that I reverted also 1a3ed0dc3594 ("mmc: sdhci-xenon: fix 1.8v regulator stabilization") commit and then issue appeared again.

Quoted 2 lines

So any of this commit is fixing that issue on Armada 3720.

I think both can stay, but if I had to choose, I'd keep 1a3ed0dc3594 ("mmc: sdhci-xenon: fix 1.8v regulator stabilization"):
* Now we know for sure it fixes both Armada 3720 and CN913x/Armada 7k8k
* Afaik this patch was checked with HW team (and IMO it looks a bit less hacky than the extra read :) ).

Best regards,
Marcin

Quoted 32 lines
Quoted 32 lines

Should we revert one of them?

Quoted 30 lines

Best regards,
Marcin

Quoted 27 lines
Quoted 27 lines
Quoted 27 lines
Quoted 27 lines
Quoted 27 lines

---
 drivers/mmc/host/sdhci-xenon.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
drivers/mmc/host/sdhci-xenon.c +6 -2
index c67611fdaa8a..4b05f6fdefb4 100644
@@ -168,7 +168,12 @@ static void xenon_reset_exit(struct sdhci_host *host,
        /* Disable tuning request and auto-retuning again */
        xenon_retune_setup(host);
 
-       xenon_set_acg(host, true);
+       /*
+        * The ACG should be turned off at the early init time, in order
+        * to solve a possile issues with the 1.8V regulator stabilization.
+        * The feature is enabled in later stage.
+        */
+       xenon_set_acg(host, false);
 
        xenon_set_sdclk_off_idle(host, sdhc_id, false);
 
--

2.29.0