Skip to contentSkip to conversation

Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

PATCH New Lukasz Majewski 3 participants 16 messages May 7, 2019 – May 20, 2019 MMC, SD and SDIO

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

This reverts commit 72a89e0da5ac6a4ab929b15a2b656f04f50767f6, which causes the imx53 HSC to hang as the eMMC is not working properly anymore.

The exact error message:
MMC write: dev # 0, block # 2, count 927 ... mmc write failed 0 blocks written: ERROR

imx53 is not using the DDR mode.

Debugging of pre_div and div generation showed that those values are generated in a way, which is not matching the ones from working setup.

As the original patch was performing code refactoring, let's revert this change, so all imx53 boards would work again.

---
 drivers/mmc/fsl_esdhc.c | 23 +++++------------------
 1 file changed, 5 insertions(+), 18 deletions(-)
drivers/mmc/fsl_esdhc.c +5 -18
index 1b7de74a72..377b2673a3 100644
@@ -621,31 +621,18 @@ static void set_sysctl(struct fsl_esdhc_priv *priv, struct mmc *mmc, uint clock)
 #else
 	int pre_div = 2;
 #endif
+	int ddr_pre_div = mmc->ddr_mode ? 2 : 1;
 	int sdhc_clk = priv->sdhc_clk;
 	uint clk;
 
-	/*
-	 * For ddr mode, usdhc need to enable DDR mode first, after select
-	 * this DDR mode, usdhc will automatically divide the usdhc clock
-	 */
-	if (mmc->ddr_mode) {
-		writel(readl(&regs->mixctrl) | MIX_CTRL_DDREN, &regs->mixctrl);
-		sdhc_clk >>= 1;
-	}
-
 	if (clock < mmc->cfg->f_min)
 		clock = mmc->cfg->f_min;
 
-	if (sdhc_clk / 16 > clock) {
-		for (; pre_div < 256; pre_div *= 2)
-			if ((sdhc_clk / pre_div) <= (clock * 16))
-				break;
-	} else
-		pre_div = 1;
+	while (sdhc_clk / (16 * pre_div * ddr_pre_div) > clock && pre_div < 256)
+		pre_div *= 2;
 
-	for (div = 1; div <= 16; div++)
-		if ((sdhc_clk / (div * pre_div)) <= clock)
-			break;
+	while (sdhc_clk / (div * pre_div * ddr_pre_div) > clock && div < 16)
+		div++;
 
 	pre_div >>= 1;
 	div -= 1;
Signature · 1 line
2.11.0
#1 lore raw

PATCH 1/1 Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Hi Lukasz,

Quoted 10 lines

Subject: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

This reverts commit 72a89e0da5ac6a4ab929b15a2b656f04f50767f6, which causes the imx53 HSC to hang as the eMMC is not working properly anymore.

The exact error message:
MMC write: dev # 0, block # 2, count 927 ... mmc write failed 0 blocks written: ERROR

Quoted 7 lines

imx53 is not using the DDR mode.

Debugging of pre_div and div generation showed that those values are generated in a way, which is not matching the ones from working setup.

As the original patch was performing code refactoring, let's revert this change, so all imx53 boards would work again.

Could you share what is the clock value for your board?

Thanks,
Peng.

Quoted 49 lines
---
 drivers/mmc/fsl_esdhc.c | 23 +++++------------------
 1 file changed, 5 insertions(+), 18 deletions(-)
drivers/mmc/fsl_esdhc.c index renamed from drivers/mmc/fsl_esdhc.c +0 -0

1b7de74a72..377b2673a3 100644

drivers/mmc/fsl_esdhc.c +0 -0
@@ -621,31 +621,18 @@ static void set_sysctl(struct fsl_esdhc_priv *priv,

struct mmc *mmc, uint clock) #else

drivers/mmc/fsl_esdhc.c +5 -19
 	int pre_div = 2;
 #endif
+	int ddr_pre_div = mmc->ddr_mode ? 2 : 1;
 	int sdhc_clk = priv->sdhc_clk;
 	uint clk;
 
-	/*
-	 * For ddr mode, usdhc need to enable DDR mode first, after select
-	 * this DDR mode, usdhc will automatically divide the usdhc clock
-	 */
-	if (mmc->ddr_mode) {
-		writel(readl(&regs->mixctrl) | MIX_CTRL_DDREN, &regs->mixctrl);
-		sdhc_clk >>= 1;
-	}
-
 	if (clock < mmc->cfg->f_min)
 		clock = mmc->cfg->f_min;
 
-	if (sdhc_clk / 16 > clock) {
-		for (; pre_div < 256; pre_div *= 2)
-			if ((sdhc_clk / pre_div) <= (clock * 16))
-				break;
-	} else
-		pre_div = 1;
+	while (sdhc_clk / (16 * pre_div * ddr_pre_div) > clock && pre_div < 256)
+		pre_div *= 2;
 
-	for (div = 1; div <= 16; div++)
-		if ((sdhc_clk / (div * pre_div)) <= clock)
-			break;
+	while (sdhc_clk / (div * pre_div * ddr_pre_div) > clock && div < 16)
+		div++;
 
 	pre_div >>= 1;
 	div -= 1;
--

2.11.0

#2 lore raw

PATCH 1/1 Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Hi Peng,

Quoted 15 lines

Hi Lukasz,

Quoted 13 lines

Subject: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

This reverts commit 72a89e0da5ac6a4ab929b15a2b656f04f50767f6, which causes the imx53 HSC to hang as the eMMC is not working properly anymore.

The exact error message:
MMC write: dev # 0, block # 2, count 927 ... mmc write failed 0 blocks written: ERROR

imx53 is not using the DDR mode.

Quoted 8 lines
Quoted 6 lines

Debugging of pre_div and div generation showed that those values are generated in a way, which is not matching the ones from working setup.

As the original patch was performing code refactoring, let's revert this change, so all imx53 boards would work again.

Could you share what is the clock value for your board?

Sure, no problem:

Working setup:
--------------

MMC:
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240
FSL_SDHC: 0
Loading Environment from MMC...
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 1 div: 1 set_sysctl: clk: 272

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 1 div: 0 set_sysctl: clk: 256

Broken:
-------
MMC:
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240
FSL_SDHC: 0
Loading Environment from MMC...
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 0 div: 3 set_sysctl: clk: 48

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 0 div: 1 set_sysctl: clk: 16

(Please also find attached patch to reproduce debug output).

Quoted 54 lines

Thanks,
Peng.

Quoted 50 lines
---
 drivers/mmc/fsl_esdhc.c | 23 +++++------------------
 1 file changed, 5 insertions(+), 18 deletions(-)
drivers/mmc/fsl_esdhc.c index renamed from drivers/mmc/fsl_esdhc.c +0 -0

1b7de74a72..377b2673a3 100644

drivers/mmc/fsl_esdhc.c +1 -2
@@ -621,31 +621,18 @@ static void set_sysctl(struct fsl_esdhc_priv
 *priv, struct mmc *mmc, uint clock)  #else
 	int pre_div = 2;
 #endif
+	int ddr_pre_div = mmc->ddr_mode ? 2 : 1;
 	int sdhc_clk = priv->sdhc_clk;
 	uint clk;
 
-	/*
-	 * For ddr mode, usdhc need to enable DDR mode first,

after select

drivers/mmc/fsl_esdhc.c +0 -1
-	 * this DDR mode, usdhc will automatically divide the

usdhc clock

drivers/mmc/fsl_esdhc.c +0 -3
-	 */
-	if (mmc->ddr_mode) {
-		writel(readl(&regs->mixctrl) | MIX_CTRL_DDREN,

&regs->mixctrl);

drivers/mmc/fsl_esdhc.c +1 -9
-		sdhc_clk >>= 1;
-	}
-
 	if (clock < mmc->cfg->f_min)
 		clock = mmc->cfg->f_min;
 
-	if (sdhc_clk / 16 > clock) {
-		for (; pre_div < 256; pre_div *= 2)
-			if ((sdhc_clk / pre_div) <= (clock * 16))
-				break;
-	} else
-		pre_div = 1;
+	while (sdhc_clk / (16 * pre_div * ddr_pre_div) > clock &&

pre_div < 256)

drivers/mmc/fsl_esdhc.c +2 -3
+		pre_div *= 2;
 
-	for (div = 1; div <= 16; div++)
-		if ((sdhc_clk / (div * pre_div)) <= clock)
-			break;
+	while (sdhc_clk / (div * pre_div * ddr_pre_div) > clock &&

div < 16)

drivers/mmc/fsl_esdhc.c +1 -0
+		div++;
Quoted 5 lines
Quoted 4 lines
drivers/mmc/fsl_esdhc.c +0 -1
 	pre_div >>= 1;
 	div -= 1;
--

2.11.0

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: lukma at denx.de -------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-WIP.patch
Type: text/x-patch
Size: 1199 bytes
Desc: not available
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20190508/a02ee699/attachment.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 488 bytes
Desc: OpenPGP digital signature
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20190508/a02ee699/attachment.sig>

#3 lore raw

PATCH 1/1 Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Quoted 64 lines · On Wed, 8 May 2019 08:19:45 +0200 Lukasz Majewski <[email protected]> wrote:

Hi Peng,

Quoted 15 lines

Hi Lukasz,

Quoted 13 lines

Subject: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

This reverts commit 72a89e0da5ac6a4ab929b15a2b656f04f50767f6, which causes the imx53 HSC to hang as the eMMC is not working properly anymore.

The exact error message:
MMC write: dev # 0, block # 2, count 927 ... mmc write failed 0 blocks written: ERROR

imx53 is not using the DDR mode.

Quoted 8 lines
Quoted 6 lines

Debugging of pre_div and div generation showed that those values are generated in a way, which is not matching the ones from working setup.

As the original patch was performing code refactoring, let's revert this change, so all imx53 boards would work again.

Could you share what is the clock value for your board?

Sure, no problem:

Working setup:
--------------

MMC:
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240
FSL_SDHC: 0
Loading Environment from MMC...
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 1 div: 1 set_sysctl: clk: 272

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 1 div: 0 set_sysctl: clk: 256

Broken:
-------
MMC:
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240
FSL_SDHC: 0
Loading Environment from MMC...
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

Quoted 9 lines

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 0 div: 3 set_sysctl: clk: 48

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 0 div: 1 set_sysctl: clk: 16

(Please also find attached patch to reproduce debug output).

And maybe the most important question - why it was necessary to refactor this code?

Parts responsible for calculating pre_div and div seems not related to ddr problem (one spot issue is div <= 16 , but in the original code it was div < 16)?

Quoted 65 lines
Quoted 54 lines

Thanks,
Peng.

Quoted 50 lines
---
 drivers/mmc/fsl_esdhc.c | 23 +++++------------------
 1 file changed, 5 insertions(+), 18 deletions(-)
drivers/mmc/fsl_esdhc.c +1 -2
index 1b7de74a72..377b2673a3 100644
@@ -621,31 +621,18 @@ static void set_sysctl(struct fsl_esdhc_priv
 *priv, struct mmc *mmc, uint clock)  #else
 	int pre_div = 2;
 #endif
+	int ddr_pre_div = mmc->ddr_mode ? 2 : 1;
 	int sdhc_clk = priv->sdhc_clk;
 	uint clk;
 
-	/*
-	 * For ddr mode, usdhc need to enable DDR mode first,

after select

drivers/mmc/fsl_esdhc.c +0 -1
-	 * this DDR mode, usdhc will automatically divide the

usdhc clock

drivers/mmc/fsl_esdhc.c +0 -3
-	 */
-	if (mmc->ddr_mode) {
-		writel(readl(&regs->mixctrl) | MIX_CTRL_DDREN,

&regs->mixctrl);

drivers/mmc/fsl_esdhc.c +1 -9
-		sdhc_clk >>= 1;
-	}
-
 	if (clock < mmc->cfg->f_min)
 		clock = mmc->cfg->f_min;
 
-	if (sdhc_clk / 16 > clock) {
-		for (; pre_div < 256; pre_div *= 2)
-			if ((sdhc_clk / pre_div) <= (clock * 16))
-				break;
-	} else
-		pre_div = 1;
+	while (sdhc_clk / (16 * pre_div * ddr_pre_div) > clock &&

pre_div < 256)

drivers/mmc/fsl_esdhc.c +2 -3
+		pre_div *= 2;
 
-	for (div = 1; div <= 16; div++)
-		if ((sdhc_clk / (div * pre_div)) <= clock)
-			break;
+	while (sdhc_clk / (div * pre_div * ddr_pre_div) > clock

&& div < 16)

drivers/mmc/fsl_esdhc.c +1 -0
+		div++;
Quoted 5 lines
Quoted 4 lines
drivers/mmc/fsl_esdhc.c +0 -1
 	pre_div >>= 1;
 	div -= 1;
--

2.11.0

Best regards,

Quoted 8 lines

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email:
lukma at denx.de

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: lukma at denx.de -------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 488 bytes
Desc: OpenPGP digital signature
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20190508/185091df/attachment.sig>

#4 lore raw

PATCH 1/1 Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Quoted 77 lines

Subject: Re: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Quoted 71 lines · On Wed, 8 May 2019 08:19:45 +0200 Lukasz Majewski <[email protected]> wrote:

Hi Peng,

Quoted 15 lines

Hi Lukasz,

Quoted 13 lines

Subject: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

This reverts commit 72a89e0da5ac6a4ab929b15a2b656f04f50767f6, which causes the imx53 HSC to hang as the eMMC is not working properly anymore.

The exact error message:
MMC write: dev # 0, block # 2, count 927 ... mmc write failed 0 blocks written: ERROR

imx53 is not using the DDR mode.

Quoted 8 lines
Quoted 6 lines

Debugging of pre_div and div generation showed that those values are generated in a way, which is not matching the ones from working setup.

As the original patch was performing code refactoring, let's revert this change, so all imx53 boards would work again.

Could you share what is the clock value for your board?

Sure, no problem:

Working setup:
--------------

MMC:
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240
FSL_SDHC: 0
Loading Environment from MMC...
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 1 div: 1 set_sysctl: clk: 272

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 1 div: 0 set_sysctl: clk: 256

Broken:
-------
MMC:
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240
FSL_SDHC: 0
Loading Environment from MMC...
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 0 div: 3 set_sysctl: clk: 48

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 0 div: 1 set_sysctl: clk: 16

Quoted 8 lines
Quoted 2 lines

(Please also find attached patch to reproduce debug output).

And maybe the most important question - why it was necessary to refactor this code?

Parts responsible for calculating pre_div and div seems not related to ddr problem (one spot issue is div <= 16 , but in the original code it was div < 16)?

This patch is mostly to support HS400 DDR_EN mode,
But after I check more, seems it is not a right fix, sorry for this regression.

It does almost the same calculation as before code expect writel(readl(&regs->mixctrl) | MIX_CTRL_DDREN, &regs->mixctrl);

and one condition break others.

So the fix for you is

drivers/mmc/fsl_esdhc.c +1 -2
index 1b7de74a72..3347fbe738 100644
@@ -640,8 +640,7 @@ static void set_sysctl(struct fsl_esdhc_priv *priv, struct mmc *mmc, uint clock)
                for (; pre_div < 256; pre_div *= 2)
                        if ((sdhc_clk / pre_div) <= (clock * 16))
                                break;
-       } else
-               pre_div = 1;
+       }
 
        for (div = 1; div <= 16; div++)
                if ((sdhc_clk / (div * pre_div)) <= clock)

With this patch, the i.MX8 HS400 will work@100MHz not 200MHz, But it not hang anyway. To fix the i.MX8, need define ARCH_MXC, I'll add a new patch for ARCH_MXC part for i.MX family.

Regards,
Peng.

Quoted 89 lines
Quoted 64 lines
Quoted 55 lines

Thanks,
Peng.

Quoted 50 lines
---
 drivers/mmc/fsl_esdhc.c | 23 +++++------------------
 1 file changed, 5 insertions(+), 18 deletions(-)
drivers/mmc/fsl_esdhc.c +1 -2
index 1b7de74a72..377b2673a3 100644
@@ -621,31 +621,18 @@ static void set_sysctl(struct fsl_esdhc_priv
 *priv, struct mmc *mmc, uint clock)  #else
 	int pre_div = 2;
 #endif
+	int ddr_pre_div = mmc->ddr_mode ? 2 : 1;
 	int sdhc_clk = priv->sdhc_clk;
 	uint clk;
 
-	/*
-	 * For ddr mode, usdhc need to enable DDR mode first,

after select

drivers/mmc/fsl_esdhc.c +0 -1
-	 * this DDR mode, usdhc will automatically divide the

usdhc clock

drivers/mmc/fsl_esdhc.c +0 -3
-	 */
-	if (mmc->ddr_mode) {
-		writel(readl(&regs->mixctrl) | MIX_CTRL_DDREN,

&regs->mixctrl);

drivers/mmc/fsl_esdhc.c +1 -9
-		sdhc_clk >>= 1;
-	}
-
 	if (clock < mmc->cfg->f_min)
 		clock = mmc->cfg->f_min;
 
-	if (sdhc_clk / 16 > clock) {
-		for (; pre_div < 256; pre_div *= 2)
-			if ((sdhc_clk / pre_div) <= (clock * 16))
-				break;
-	} else
-		pre_div = 1;
+	while (sdhc_clk / (16 * pre_div * ddr_pre_div) > clock &&

pre_div < 256)

drivers/mmc/fsl_esdhc.c +2 -3
+		pre_div *= 2;
 
-	for (div = 1; div <= 16; div++)
-		if ((sdhc_clk / (div * pre_div)) <= clock)
-			break;
+	while (sdhc_clk / (div * pre_div * ddr_pre_div) > clock

&& div < 16)

drivers/mmc/fsl_esdhc.c +1 -0
+		div++;
Quoted 5 lines
Quoted 4 lines
drivers/mmc/fsl_esdhc.c +0 -1
 	pre_div >>= 1;
 	div -= 1;
--

2.11.0

Quoted 7 lines

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang

Denk

Quoted 3 lines

HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email:
lukma at denx.de

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email:
lukma at denx.de

#5 lore raw

PATCH 1/1 Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Hi Peng,

Quoted 113 lines
Quoted 80 lines

Subject: Re: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Quoted 64 lines · On Wed, 8 May 2019 08:19:45 +0200 Lukasz Majewski <[email protected]> wrote:

Hi Peng,

Quoted 15 lines

Hi Lukasz,

Quoted 13 lines

Subject: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

This reverts commit 72a89e0da5ac6a4ab929b15a2b656f04f50767f6, which causes the imx53 HSC to hang as the eMMC is not working properly anymore.

The exact error message:
MMC write: dev # 0, block # 2, count 927 ... mmc write failed 0 blocks written: ERROR

imx53 is not using the DDR mode.

Quoted 8 lines
Quoted 6 lines

Debugging of pre_div and div generation showed that those values are generated in a way, which is not matching the ones from working setup.

As the original patch was performing code refactoring, let's revert this change, so all imx53 boards would work again.

Could you share what is the clock value for your board?

Sure, no problem:

Working setup:
--------------

MMC:
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240
FSL_SDHC: 0
Loading Environment from MMC...
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 1 div: 1 set_sysctl: clk: 272

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 1 div: 0 set_sysctl: clk: 256

Broken:
-------
MMC:
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240
FSL_SDHC: 0
Loading Environment from MMC...
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

Quoted 9 lines

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 0 div: 3 set_sysctl: clk: 48

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 0 div: 1 set_sysctl: clk: 16

(Please also find attached patch to reproduce debug output).

Quoted 6 lines

And maybe the most important question - why it was necessary to refactor this code?

Parts responsible for calculating pre_div and div seems not related to ddr problem (one spot issue is div <= 16 , but in the original code it was div < 16)?

This patch is mostly to support HS400 DDR_EN mode,
But after I check more, seems it is not a right fix, sorry for this regression.

It does almost the same calculation as before code expect writel(readl(&regs->mixctrl) | MIX_CTRL_DDREN, &regs->mixctrl);

and one condition break others.

So the fix for you is

drivers/mmc/fsl_esdhc.c +1 -2
index 1b7de74a72..3347fbe738 100644
@@ -640,8 +640,7 @@ static void set_sysctl(struct fsl_esdhc_priv
 *priv, struct mmc *mmc, uint clock) for (; pre_div < 256; pre_div *=
 2) if ((sdhc_clk / pre_div) <= (clock * 16))
                                break;
-       } else
-               pre_div = 1;
+       }
 
        for (div = 1; div <= 16; div++)
                if ((sdhc_clk / (div * pre_div)) <= clock)
Quoted 4 lines

With this patch, the i.MX8 HS400 will work at 100MHz not 200MHz, But it not hang anyway. To fix the i.MX8, need define ARCH_MXC, I'll add a new patch for ARCH_MXC part for i.MX family.

Thanks for the explanation.

I'm just wondering if we shall revert the original patch and wait for proper fix for IMX8 or will you prepare a separate patch to fix the current mainline (so the "revert" patch can be dropped) ?

Quoted 93 lines

Regards,
Peng.

Quoted 90 lines
Quoted 65 lines
Quoted 56 lines

Thanks,
Peng.

Quoted 51 lines
---
 drivers/mmc/fsl_esdhc.c | 23 +++++------------------
 1 file changed, 5 insertions(+), 18 deletions(-)
drivers/mmc/fsl_esdhc.c +0 -0
index 1b7de74a72..377b2673a3 100644
@@ -621,31 +621,18 @@ static void set_sysctl(struct

fsl_esdhc_priv *priv, struct mmc *mmc, uint clock) #else

drivers/mmc/fsl_esdhc.c +1 -2
 	int pre_div = 2;
 #endif
+	int ddr_pre_div = mmc->ddr_mode ? 2 : 1;
 	int sdhc_clk = priv->sdhc_clk;
 	uint clk;
 
-	/*
-	 * For ddr mode, usdhc need to enable DDR mode first,

after select

drivers/mmc/fsl_esdhc.c +0 -1
-	 * this DDR mode, usdhc will automatically divide the

usdhc clock

drivers/mmc/fsl_esdhc.c +0 -3
-	 */
-	if (mmc->ddr_mode) {
-		writel(readl(&regs->mixctrl) |

MIX_CTRL_DDREN, &regs->mixctrl);

drivers/mmc/fsl_esdhc.c +0 -6
-		sdhc_clk >>= 1;
-	}
-
 	if (clock < mmc->cfg->f_min)
 		clock = mmc->cfg->f_min;
 
-	if (sdhc_clk / 16 > clock) {
-		for (; pre_div < 256; pre_div *= 2)
-			if ((sdhc_clk / pre_div) <= (clock *

16))

drivers/mmc/fsl_esdhc.c +1 -3
-				break;
-	} else
-		pre_div = 1;
+	while (sdhc_clk / (16 * pre_div * ddr_pre_div) >

clock && pre_div < 256)

drivers/mmc/fsl_esdhc.c +2 -3
+		pre_div *= 2;
 
-	for (div = 1; div <= 16; div++)
-		if ((sdhc_clk / (div * pre_div)) <= clock)
-			break;
+	while (sdhc_clk / (div * pre_div * ddr_pre_div) >

clock && div < 16)

drivers/mmc/fsl_esdhc.c +1 -0
+		div++;
Quoted 5 lines
Quoted 4 lines
drivers/mmc/fsl_esdhc.c +0 -1
 	pre_div >>= 1;
 	div -= 1;
--

2.11.0

Quoted 7 lines

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang

Denk

Quoted 3 lines

HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email:
lukma at denx.de

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang
Denk HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email:
lukma at denx.de

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: lukma at denx.de -------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 488 bytes
Desc: OpenPGP digital signature
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20190508/11abbade/attachment.sig>

#6 lore raw

PATCH 1/1 Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Quoted 134 lines

-----Original Message-----
From: Lukasz Majewski [mailto:lukma at denx.de]
Sent: 2019年5月8日 19:38
To: Peng Fan <[email protected]>
Cc: u-boot at lists.denx.de; Tom Rini <[email protected]>; Marcel Ziswiler <[email protected]>; Fabio Estevam <[email protected]>; dl-uboot-imx <[email protected]>; sbabic at denx.de; Stefan Agner <[email protected]>; BOUGH CHEN <[email protected]>; Ye Li <[email protected]>
Subject: Re: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Hi Peng,

Quoted 112 lines
Quoted 80 lines

Subject: Re: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Quoted 64 lines · On Wed, 8 May 2019 08:19:45 +0200 Lukasz Majewski <[email protected]> wrote:

Hi Peng,

Quoted 15 lines

Hi Lukasz,

Quoted 13 lines

Subject: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

This reverts commit 72a89e0da5ac6a4ab929b15a2b656f04f50767f6, which causes the imx53 HSC to hang as the eMMC is not working properly anymore.

The exact error message:
MMC write: dev # 0, block # 2, count 927 ... mmc write failed 0 blocks written: ERROR

imx53 is not using the DDR mode.

Quoted 8 lines
Quoted 6 lines

Debugging of pre_div and div generation showed that those values are generated in a way, which is not matching the ones from working setup.

As the original patch was performing code refactoring, let's revert this change, so all imx53 boards would work again.

Could you share what is the clock value for your board?

Sure, no problem:

Working setup:
--------------

MMC:
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240
FSL_SDHC: 0
Loading Environment from MMC...
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 1 div: 1 set_sysctl: clk: 272

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 1 div: 0 set_sysctl: clk: 256

Broken:
-------
MMC:
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240
FSL_SDHC: 0
Loading Environment from MMC...
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

Quoted 9 lines

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 0 div: 3 set_sysctl: clk: 48

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 0 div: 1 set_sysctl: clk: 16

(Please also find attached patch to reproduce debug output).

Quoted 6 lines

And maybe the most important question - why it was necessary to refactor this code?

Parts responsible for calculating pre_div and div seems not related to ddr problem (one spot issue is div <= 16 , but in the original code it was div < 16)?

This patch is mostly to support HS400 DDR_EN mode, But after I check more, seems it is not a right fix, sorry for this regression.

It does almost the same calculation as before code expect writel(readl(&regs->mixctrl) | MIX_CTRL_DDREN, &regs->mixctrl);

and one condition break others.

So the fix for you is

drivers/mmc/fsl_esdhc.c index renamed from drivers/mmc/fsl_esdhc.c +0 -0

1b7de74a72..3347fbe738 100644

drivers/mmc/fsl_esdhc.c +1 -2
@@ -640,8 +640,7 @@ static void set_sysctl(struct fsl_esdhc_priv
 *priv, struct mmc *mmc, uint clock) for (; pre_div < 256; pre_div *=
 2) if ((sdhc_clk / pre_div) <= (clock * 16))
                                break;
-       } else
-               pre_div = 1;
+       }
 
        for (div = 1; div <= 16; div++)
                if ((sdhc_clk / (div * pre_div)) <= clock)
Quoted 4 lines

With this patch, the i.MX8 HS400 will work at 100MHz not 200MHz, But it not hang anyway. To fix the i.MX8, need define ARCH_MXC, I'll add a new patch for ARCH_MXC part for i.MX family.

Thanks for the explanation.

Quoted 3 lines

I'm just wondering if we shall revert the original patch and wait for proper fix for IMX8 or will you prepare a separate patch to fix the current mainline (so the "revert" patch can be dropped) ?

When we add the patch, we do not take the i.mx53 case into consideration. Sorry for this mistake. I prefer add a fix patch rather than 'revert', because I notice there is a claim in the RM:

If setting SDCLKFS and DVS can generate same clock frequency,(For example, in SDR mode, SDCLKFS = 01h is same as DVS = 01h.) SDCLKFS is highly recommended.

For both these methods, the current one or the previous one, the code prefer DVS rather than SDCLKFS, not align with the RM suggestion. So this time I plan to re-write the code to follow the RM suggestion. I will send a patch these week.

Regards,
Bough Chen

Quoted 107 lines
Quoted 94 lines

Regards,
Peng.

Quoted 90 lines
Quoted 65 lines
Quoted 56 lines

Thanks,
Peng.

Quoted 51 lines
---
 drivers/mmc/fsl_esdhc.c | 23 +++++------------------
 1 file changed, 5 insertions(+), 18 deletions(-)
drivers/mmc/fsl_esdhc.c +0 -0
index 1b7de74a72..377b2673a3 100644
@@ -621,31 +621,18 @@ static void set_sysctl(struct

fsl_esdhc_priv *priv, struct mmc *mmc, uint clock) #else

drivers/mmc/fsl_esdhc.c +1 -2
 	int pre_div = 2;
 #endif
+	int ddr_pre_div = mmc->ddr_mode ? 2 : 1;
 	int sdhc_clk = priv->sdhc_clk;
 	uint clk;
 
-	/*
-	 * For ddr mode, usdhc need to enable DDR mode first,

after select

drivers/mmc/fsl_esdhc.c +0 -1
-	 * this DDR mode, usdhc will automatically divide the

usdhc clock

drivers/mmc/fsl_esdhc.c +0 -3
-	 */
-	if (mmc->ddr_mode) {
-		writel(readl(&regs->mixctrl) |

MIX_CTRL_DDREN, &regs->mixctrl);

drivers/mmc/fsl_esdhc.c +0 -6
-		sdhc_clk >>= 1;
-	}
-
 	if (clock < mmc->cfg->f_min)
 		clock = mmc->cfg->f_min;
 
-	if (sdhc_clk / 16 > clock) {
-		for (; pre_div < 256; pre_div *= 2)
-			if ((sdhc_clk / pre_div) <= (clock *

16))

drivers/mmc/fsl_esdhc.c +1 -3
-				break;
-	} else
-		pre_div = 1;
+	while (sdhc_clk / (16 * pre_div * ddr_pre_div) >

clock && pre_div < 256)

drivers/mmc/fsl_esdhc.c +2 -3
+		pre_div *= 2;
 
-	for (div = 1; div <= 16; div++)
-		if ((sdhc_clk / (div * pre_div)) <= clock)
-			break;
+	while (sdhc_clk / (div * pre_div * ddr_pre_div) >

clock && div < 16)

drivers/mmc/fsl_esdhc.c +1 -0
+		div++;
Quoted 5 lines
Quoted 4 lines
drivers/mmc/fsl_esdhc.c +0 -1
 	pre_div >>= 1;
 	div -= 1;
--

2.11.0

Quoted 7 lines

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang

Denk

Quoted 3 lines

HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email:
lukma at denx.de

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang
Denk HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email:
lukma at denx.de

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: lukma at denx.de

#7 lore raw

PATCH 1/1 Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Quoted 144 lines · On Wed, 8 May 2019 13:03:58 +0000 BOUGH CHEN <[email protected]> wrote:
Quoted 136 lines

-----Original Message-----
From: Lukasz Majewski [mailto:lukma at denx.de]
Sent: 2019年5月8日 19:38
To: Peng Fan <[email protected]>
Cc: u-boot at lists.denx.de; Tom Rini <[email protected]>; Marcel Ziswiler <[email protected]>; Fabio Estevam <[email protected]>; dl-uboot-imx <[email protected]>; sbabic at denx.de; Stefan Agner <[email protected]>; BOUGH CHEN <[email protected]>; Ye Li <[email protected]>
Subject: Re: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Hi Peng,

Quoted 115 lines
Quoted 82 lines

Subject: Re: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Quoted 66 lines · On Wed, 8 May 2019 08:19:45 +0200 Lukasz Majewski <[email protected]> wrote:

Hi Peng,

Quoted 20 lines

Hi Lukasz,

Quoted 18 lines

Subject: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

This reverts commit 72a89e0da5ac6a4ab929b15a2b656f04f50767f6, which causes the imx53 HSC to hang as the eMMC is not working properly anymore.

The exact error message:
MMC write: dev # 0, block # 2, count 927 ... mmc write failed 0 blocks written: ERROR

imx53 is not using the DDR mode.

Debugging of pre_div and div generation showed that those values are generated in a way, which is not matching the ones from working setup.

Quoted 5 lines
Quoted 3 lines

As the original patch was performing code refactoring, let's revert this change, so all imx53 boards would work again.

Could you share what is the clock value for your board?

Sure, no problem:

Working setup:
--------------

MMC:
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240
FSL_SDHC: 0
Loading Environment from MMC...
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 1 div: 1 set_sysctl: clk: 272

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 1 div: 0 set_sysctl: clk: 256

Broken:
-------
MMC:
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240
FSL_SDHC: 0
Loading Environment from MMC...
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

Quoted 9 lines

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 0 div: 3 set_sysctl: clk: 48

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 0 div: 1 set_sysctl: clk: 16

(Please also find attached patch to reproduce debug output).

Quoted 6 lines

And maybe the most important question - why it was necessary to refactor this code?

Parts responsible for calculating pre_div and div seems not related to ddr problem (one spot issue is div <= 16 , but in the original code it was div < 16)?

This patch is mostly to support HS400 DDR_EN mode, But after I check more, seems it is not a right fix, sorry for this regression.

It does almost the same calculation as before code expect writel(readl(&regs->mixctrl) | MIX_CTRL_DDREN, &regs->mixctrl);

and one condition break others.

So the fix for you is

drivers/mmc/fsl_esdhc.c +1 -2
index 1b7de74a72..3347fbe738 100644
@@ -640,8 +640,7 @@ static void set_sysctl(struct fsl_esdhc_priv
 *priv, struct mmc *mmc, uint clock) for (; pre_div < 256; pre_div
 *= 2) if ((sdhc_clk / pre_div) <= (clock * 16))
                                break;
-       } else
-               pre_div = 1;
+       }
 
        for (div = 1; div <= 16; div++)
                if ((sdhc_clk / (div * pre_div)) <= clock)
Quoted 4 lines

With this patch, the i.MX8 HS400 will work at 100MHz not 200MHz, But it not hang anyway. To fix the i.MX8, need define ARCH_MXC, I'll add a new patch for ARCH_MXC part for i.MX family.

Thanks for the explanation.

Quoted 3 lines

I'm just wondering if we shall revert the original patch and wait for proper fix for IMX8 or will you prepare a separate patch to fix the current mainline (so the "revert" patch can be dropped) ?

When we add the patch, we do not take the i.mx53 case into consideration. Sorry for this mistake. I prefer add a fix patch rather than 'revert', because I notice there is a claim in the RM:

Quoted 9 lines

If setting SDCLKFS and DVS can generate same clock frequency,(For
example, in SDR mode, SDCLKFS = 01h is same as DVS = 01h.) SDCLKFS is highly recommended.

For both these methods, the current one or the previous one, the code prefer DVS rather than SDCLKFS, not align with the RM suggestion. So this time I plan to re-write the code to follow the RM suggestion. I will send a patch these week.

Ok. Thanks for the update.

I'm looking forward to test your patch.

Quoted 113 lines

Regards,
Bough Chen

Quoted 110 lines
Quoted 96 lines

Regards,
Peng.

Quoted 92 lines
Quoted 66 lines
Quoted 57 lines

Thanks,
Peng.

Quoted 52 lines
---
 drivers/mmc/fsl_esdhc.c | 23 +++++------------------
 1 file changed, 5 insertions(+), 18 deletions(-)
(quoted diff) +0 -0

b/drivers/mmc/fsl_esdhc.c index 1b7de74a72..377b2673a3 100644 --- a/drivers/mmc/fsl_esdhc.c
+++ b/drivers/mmc/fsl_esdhc.c

(quoted diff) +0 -0
@@ -621,31 +621,18 @@ static void set_sysctl(struct

fsl_esdhc_priv *priv, struct mmc *mmc, uint clock) #else

(quoted diff) +1 -2
 	int pre_div = 2;
 #endif
+	int ddr_pre_div = mmc->ddr_mode ? 2 : 1;
 	int sdhc_clk = priv->sdhc_clk;
 	uint clk;
 
-	/*
-	 * For ddr mode, usdhc need to enable DDR mode

first, after select

(quoted diff) +0 -1
-	 * this DDR mode, usdhc will automatically

divide the usdhc clock

(quoted diff) +0 -3
-	 */
-	if (mmc->ddr_mode) {
-		writel(readl(&regs->mixctrl) |

MIX_CTRL_DDREN, &regs->mixctrl);

(quoted diff) +0 -6
-		sdhc_clk >>= 1;
-	}
-
 	if (clock < mmc->cfg->f_min)
 		clock = mmc->cfg->f_min;
 
-	if (sdhc_clk / 16 > clock) {
-		for (; pre_div < 256; pre_div *= 2)
-			if ((sdhc_clk / pre_div) <=

(clock * 16))

drivers/mmc/fsl_esdhc.c +1 -3
-				break;
-	} else
-		pre_div = 1;
+	while (sdhc_clk / (16 * pre_div * ddr_pre_div) >

clock && pre_div < 256)

(quoted diff) +1 -2
+		pre_div *= 2;
 
-	for (div = 1; div <= 16; div++)
-		if ((sdhc_clk / (div * pre_div)) <=

clock)

(quoted diff) +1 -1
-			break;
+	while (sdhc_clk / (div * pre_div * ddr_pre_div) >

clock && div < 16)

(quoted diff) +1 -0
+		div++;
Quoted 5 lines
Quoted 4 lines
(quoted diff) +0 -1
 	pre_div >>= 1;
 	div -= 1;
--

2.11.0

Quoted 8 lines

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director:
Wolfgang

Denk

Quoted 3 lines

HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80
Email: lukma at denx.de

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang
Denk HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-59 Fax:
(+49)-8142-66989-80 Email: lukma at denx.de

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang
Denk HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email:
lukma at denx.de

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: lukma at denx.de -------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 488 bytes
Desc: OpenPGP digital signature
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20190508/ef550c93/attachment.sig>

#8 lore raw

PATCH 1/1 Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

In reply to Lukasz Majewski
Quoted 88 lines

-----Original Message-----
From: Lukasz Majewski [mailto:lukma at denx.de]
Sent: 2019年5月8日 14:38
To: Peng Fan <[email protected]>
Cc: u-boot at lists.denx.de; Tom Rini <[email protected]>; Marcel Ziswiler <[email protected]>; Fabio Estevam <[email protected]>; dl-uboot-imx <[email protected]>; sbabic at denx.de; Stefan Agner <[email protected]>; BOUGH CHEN <[email protected]>; Ye Li <[email protected]>
Subject: Re: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Quoted 71 lines · On Wed, 8 May 2019 08:19:45 +0200 Lukasz Majewski <[email protected]> wrote:

Hi Peng,

Quoted 15 lines

Hi Lukasz,

Quoted 13 lines

Subject: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

This reverts commit 72a89e0da5ac6a4ab929b15a2b656f04f50767f6, which causes the imx53 HSC to hang as the eMMC is not working properly anymore.

The exact error message:
MMC write: dev # 0, block # 2, count 927 ... mmc write failed 0 blocks written: ERROR

imx53 is not using the DDR mode.

Quoted 8 lines
Quoted 6 lines

Debugging of pre_div and div generation showed that those values are generated in a way, which is not matching the ones from working setup.

As the original patch was performing code refactoring, let's revert this change, so all imx53 boards would work again.

Could you share what is the clock value for your board?

Sure, no problem:

Working setup:
--------------

MMC:
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240
FSL_SDHC: 0
Loading Environment from MMC...
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 1 div: 1 set_sysctl: clk: 272

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 1 div: 0 set_sysctl: clk: 256

Broken:
-------
MMC:
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240
FSL_SDHC: 0
Loading Environment from MMC...
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 0 div: 3 set_sysctl: clk: 48

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 0 div: 1 set_sysctl: clk: 16

Quoted 8 lines
Quoted 2 lines

(Please also find attached patch to reproduce debug output).

And maybe the most important question - why it was necessary to refactor this code?

Parts responsible for calculating pre_div and div seems not related to ddr problem (one spot issue is div <= 16 , but in the original code it was div < 16)?

Could you help verify whether the patch fixes you issue?

index 1b7de74a72..3347fbe738 100644

drivers/mmc/fsl_esdhc.c +1 -2
@@ -640,8 +640,7 @@ static void set_sysctl(struct fsl_esdhc_priv *priv, struct mmc *mmc, uint clock)
                for (; pre_div < 256; pre_div *= 2)
                        if ((sdhc_clk / pre_div) <= (clock * 16))
                                break;
-       } else
-               pre_div = 1;
+       }
 
        for (div = 1; div <= 16; div++)
                if ((sdhc_clk / (div * pre_div)) <= clock)

Thanks,
Peng.

Quoted 89 lines
Quoted 64 lines
Quoted 55 lines

Thanks,
Peng.

Quoted 50 lines
---
 drivers/mmc/fsl_esdhc.c | 23 +++++------------------
 1 file changed, 5 insertions(+), 18 deletions(-)
drivers/mmc/fsl_esdhc.c +1 -2
index 1b7de74a72..377b2673a3 100644
@@ -621,31 +621,18 @@ static void set_sysctl(struct fsl_esdhc_priv
 *priv, struct mmc *mmc, uint clock)  #else
 	int pre_div = 2;
 #endif
+	int ddr_pre_div = mmc->ddr_mode ? 2 : 1;
 	int sdhc_clk = priv->sdhc_clk;
 	uint clk;
 
-	/*
-	 * For ddr mode, usdhc need to enable DDR mode first,

after select

drivers/mmc/fsl_esdhc.c +0 -1
-	 * this DDR mode, usdhc will automatically divide the

usdhc clock

drivers/mmc/fsl_esdhc.c +0 -3
-	 */
-	if (mmc->ddr_mode) {
-		writel(readl(&regs->mixctrl) | MIX_CTRL_DDREN,

&regs->mixctrl);

drivers/mmc/fsl_esdhc.c +1 -9
-		sdhc_clk >>= 1;
-	}
-
 	if (clock < mmc->cfg->f_min)
 		clock = mmc->cfg->f_min;
 
-	if (sdhc_clk / 16 > clock) {
-		for (; pre_div < 256; pre_div *= 2)
-			if ((sdhc_clk / pre_div) <= (clock * 16))
-				break;
-	} else
-		pre_div = 1;
+	while (sdhc_clk / (16 * pre_div * ddr_pre_div) > clock &&

pre_div < 256)

drivers/mmc/fsl_esdhc.c +2 -3
+		pre_div *= 2;
 
-	for (div = 1; div <= 16; div++)
-		if ((sdhc_clk / (div * pre_div)) <= clock)
-			break;
+	while (sdhc_clk / (div * pre_div * ddr_pre_div) > clock

&& div < 16)

drivers/mmc/fsl_esdhc.c +1 -0
+		div++;
Quoted 5 lines
Quoted 4 lines
drivers/mmc/fsl_esdhc.c +0 -1
 	pre_div >>= 1;
 	div -= 1;
--

2.11.0

Quoted 7 lines

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang

Denk

Quoted 3 lines

HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email:
lukma at denx.de

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email:
lukma at denx.de

#9 lore raw

PATCH 1/1 Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Quoted 102 lines · On Wed, 8 May 2019 13:59:14 +0000 Peng Fan <[email protected]> wrote:
Quoted 89 lines

-----Original Message-----
From: Lukasz Majewski [mailto:lukma at denx.de]
Sent: 2019年5月8日 14:38
To: Peng Fan <[email protected]>
Cc: u-boot at lists.denx.de; Tom Rini <[email protected]>; Marcel Ziswiler <[email protected]>; Fabio Estevam <[email protected]>; dl-uboot-imx <[email protected]>; sbabic at denx.de; Stefan Agner <[email protected]>; BOUGH CHEN <[email protected]>; Ye Li <[email protected]>
Subject: Re: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Quoted 64 lines · On Wed, 8 May 2019 08:19:45 +0200 Lukasz Majewski <[email protected]> wrote:

Hi Peng,

Quoted 15 lines

Hi Lukasz,

Quoted 13 lines

Subject: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

This reverts commit 72a89e0da5ac6a4ab929b15a2b656f04f50767f6, which causes the imx53 HSC to hang as the eMMC is not working properly anymore.

The exact error message:
MMC write: dev # 0, block # 2, count 927 ... mmc write failed 0 blocks written: ERROR

imx53 is not using the DDR mode.

Quoted 8 lines
Quoted 6 lines

Debugging of pre_div and div generation showed that those values are generated in a way, which is not matching the ones from working setup.

As the original patch was performing code refactoring, let's revert this change, so all imx53 boards would work again.

Could you share what is the clock value for your board?

Sure, no problem:

Working setup:
--------------

MMC:
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240
FSL_SDHC: 0
Loading Environment from MMC...
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 1 div: 1 set_sysctl: clk: 272

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 1 div: 0 set_sysctl: clk: 256

Broken:
-------
MMC:
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240
FSL_SDHC: 0
Loading Environment from MMC...
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

Quoted 9 lines

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 0 div: 3 set_sysctl: clk: 48

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 0 div: 1 set_sysctl: clk: 16

(Please also find attached patch to reproduce debug output).

Quoted 6 lines

And maybe the most important question - why it was necessary to refactor this code?

Parts responsible for calculating pre_div and div seems not related to ddr problem (one spot issue is div <= 16 , but in the original code it was div < 16)?

Could you help verify whether the patch fixes you issue?

index 1b7de74a72..3347fbe738 100644

drivers/mmc/fsl_esdhc.c static void set_sysctl(struct fsl_esdhc_priv · line 643

Quoted 6 lines
drivers/mmc/fsl_esdhc.c +1 -2
@@ -640,8 +640,7 @@ static void set_sysctl(struct fsl_esdhc_priv
 *priv, struct mmc *mmc, uint clock) for (; pre_div < 256; pre_div *=
 2) if ((sdhc_clk / pre_div) <= (clock * 16))
                                break;
-       } else
-               pre_div = 1;
+       }

Please examine this code thoroughly and provide patch.

The
} else

	pre_div = 1;

was added there for a purpose, so I'm wondering why it can be easily removed now.

drivers/mmc/fsl_esdhc.c static void set_sysctl(struct fsl_esdhc_priv *priv, struct mmc *mmc, uint clock)

Quoted 4 lines
drivers/mmc/fsl_esdhc.c +0 -0
        for (div = 1; div <= 16; div++)
                if ((sdhc_clk / (div * pre_div)) <= clock)
 

As I've stated above - is the above for() correct?

In the original code it was div < 16, but here it is div <= 16.

Quoted 94 lines

Thanks,
Peng.

Quoted 90 lines
Quoted 65 lines
Quoted 56 lines

Thanks,
Peng.

Quoted 51 lines
---
 drivers/mmc/fsl_esdhc.c | 23 +++++------------------
 1 file changed, 5 insertions(+), 18 deletions(-)
drivers/mmc/fsl_esdhc.c +0 -0
index 1b7de74a72..377b2673a3 100644
@@ -621,31 +621,18 @@ static void set_sysctl(struct

fsl_esdhc_priv *priv, struct mmc *mmc, uint clock) #else

drivers/mmc/fsl_esdhc.c +1 -2
 	int pre_div = 2;
 #endif
+	int ddr_pre_div = mmc->ddr_mode ? 2 : 1;
 	int sdhc_clk = priv->sdhc_clk;
 	uint clk;
 
-	/*
-	 * For ddr mode, usdhc need to enable DDR mode first,

after select

drivers/mmc/fsl_esdhc.c +0 -1
-	 * this DDR mode, usdhc will automatically divide the

usdhc clock

drivers/mmc/fsl_esdhc.c +0 -3
-	 */
-	if (mmc->ddr_mode) {
-		writel(readl(&regs->mixctrl) |

MIX_CTRL_DDREN, &regs->mixctrl);

drivers/mmc/fsl_esdhc.c +0 -6
-		sdhc_clk >>= 1;
-	}
-
 	if (clock < mmc->cfg->f_min)
 		clock = mmc->cfg->f_min;
 
-	if (sdhc_clk / 16 > clock) {
-		for (; pre_div < 256; pre_div *= 2)
-			if ((sdhc_clk / pre_div) <= (clock *

16))

drivers/mmc/fsl_esdhc.c +1 -3
-				break;
-	} else
-		pre_div = 1;
+	while (sdhc_clk / (16 * pre_div * ddr_pre_div) >

clock && pre_div < 256)

drivers/mmc/fsl_esdhc.c +2 -3
+		pre_div *= 2;
 
-	for (div = 1; div <= 16; div++)
-		if ((sdhc_clk / (div * pre_div)) <= clock)
-			break;
+	while (sdhc_clk / (div * pre_div * ddr_pre_div) >

clock && div < 16)

drivers/mmc/fsl_esdhc.c +1 -0
+		div++;
Quoted 5 lines
Quoted 4 lines
drivers/mmc/fsl_esdhc.c +0 -1
 	pre_div >>= 1;
 	div -= 1;
--

2.11.0

Quoted 7 lines

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang

Denk

Quoted 3 lines

HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email:
lukma at denx.de

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang
Denk HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email:
lukma at denx.de

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: lukma at denx.de -------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 488 bytes
Desc: OpenPGP digital signature
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20190508/1466773c/attachment.sig>

#10 lore raw

PATCH 1/1 Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Quoted 106 lines

Subject: Re: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Quoted 8 lines · On Wed, 8 May 2019 13:59:14 +0000 Peng Fan <[email protected]> wrote:
Quoted 8 lines

-----Original Message-----
From: Lukasz Majewski [mailto:lukma at denx.de]
Sent: 2019年5月8日 14:38
To: Peng Fan <[email protected]>
Cc: u-boot at lists.denx.de; Tom Rini <[email protected]>; Marcel Ziswiler <[email protected]>; Fabio Estevam <[email protected]>; dl-uboot-imx <[email protected]>; sbabic at denx.de; Stefan Agner <[email protected]>; BOUGH

CHEN

Quoted 7 lines
Quoted 7 lines

<[email protected]>; Ye Li <[email protected]>
Subject: Re: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

On Wed, 8 May 2019 08:19:45 +0200
Lukasz Majewski <[email protected]> wrote:

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines

Hi Peng,

Quoted 6 lines

Hi Lukasz,

Quoted 4 lines

Subject: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

This reverts commit

72a89e0da5ac6a4ab929b15a2b656f04f50767f6,

Quoted 75 lines
Quoted 66 lines
Quoted 56 lines
Quoted 9 lines
Quoted 9 lines

which causes the imx53 HSC to hang as the eMMC is not working properly anymore.

The exact error message:
MMC write: dev # 0, block # 2, count 927 ... mmc write failed 0 blocks written: ERROR

imx53 is not using the DDR mode.

Quoted 8 lines
Quoted 6 lines

Debugging of pre_div and div generation showed that those values are generated in a way, which is not matching the ones from working setup.

As the original patch was performing code refactoring, let's revert this change, so all imx53 boards would work again.

Could you share what is the clock value for your board?

Sure, no problem:

Working setup:
--------------

MMC:
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240
FSL_SDHC: 0
Loading Environment from MMC...
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 1 div: 1 set_sysctl: clk: 272

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 1 div: 0 set_sysctl: clk: 256

Broken:
-------
MMC:
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240
FSL_SDHC: 0
Loading Environment from MMC...
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

Quoted 9 lines

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 0 div: 3 set_sysctl: clk: 48

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 0 div: 1 set_sysctl: clk: 16

(Please also find attached patch to reproduce debug output).

Quoted 6 lines

And maybe the most important question - why it was necessary to refactor this code?

Parts responsible for calculating pre_div and div seems not related to ddr problem (one spot issue is div <= 16 , but in the original code it was div < 16)?

Could you help verify whether the patch fixes you issue?

Quoted 12 lines

drivers/mmc/fsl_esdhc.c static void set_sysctl(struct fsl_esdhc_priv · line 643

Quoted 10 lines

index 1b7de74a72..3347fbe738 100644

drivers/mmc/fsl_esdhc.c +1 -2
@@ -640,8 +640,7 @@ static void set_sysctl(struct fsl_esdhc_priv
 *priv, struct mmc *mmc, uint clock) for (; pre_div < 256; pre_div *=
 2) if ((sdhc_clk / pre_div) <= (clock * 16))
                                break;
-       } else
-               pre_div = 1;
+       }

Please examine this code thoroughly and provide patch.

The pre_div should not override the initialization value at the beginning of the function.

Quoted 7 lines

The
} else

	pre_div = 1;

was added there for a purpose, so I'm wondering why it can be easily removed now.

The else was wrongly added. It is not correct.

Quoted 9 lines

drivers/mmc/fsl_esdhc.c static void set_sysctl(struct fsl_esdhc_priv *priv, struct mmc *mmc, uint clock)

Quoted 5 lines
drivers/mmc/fsl_esdhc.c +0 -0
        for (div = 1; div <= 16; div++)
                if ((sdhc_clk / (div * pre_div)) <= clock)
 

As I've stated above - is the above for() correct?

In the original code it was div < 16, but here it is div <= 16.

Checking i.MX53 SDHC DVS, it supports [1,16], so should use "<=16", other i.MX has same.

Divisor:
This register is used to provide a more exact divisor to generate the desired SD clock frequency.
NOTE: The divider can even support odd divisor without deterioration of duty cycle. The setting are as following:
0x0 Divisor by 1
0x1 Divisor by 2
0xe Divisor by 15
0xf Divisor by 16

Regards,
Peng.

Quoted 109 lines
Quoted 95 lines

Thanks,
Peng.

Quoted 90 lines
Quoted 65 lines
Quoted 56 lines

Thanks,
Peng.

Quoted 51 lines
---
 drivers/mmc/fsl_esdhc.c | 23 +++++------------------
 1 file changed, 5 insertions(+), 18 deletions(-)
drivers/mmc/fsl_esdhc.c +0 -0
index 1b7de74a72..377b2673a3 100644
@@ -621,31 +621,18 @@ static void set_sysctl(struct

fsl_esdhc_priv *priv, struct mmc *mmc, uint clock) #else

drivers/mmc/fsl_esdhc.c +1 -2
 	int pre_div = 2;
 #endif
+	int ddr_pre_div = mmc->ddr_mode ? 2 : 1;
 	int sdhc_clk = priv->sdhc_clk;
 	uint clk;
 
-	/*
-	 * For ddr mode, usdhc need to enable DDR mode first,

after select

drivers/mmc/fsl_esdhc.c +0 -1
-	 * this DDR mode, usdhc will automatically divide the

usdhc clock

drivers/mmc/fsl_esdhc.c +0 -3
-	 */
-	if (mmc->ddr_mode) {
-		writel(readl(&regs->mixctrl) |

MIX_CTRL_DDREN, &regs->mixctrl);

drivers/mmc/fsl_esdhc.c +0 -6
-		sdhc_clk >>= 1;
-	}
-
 	if (clock < mmc->cfg->f_min)
 		clock = mmc->cfg->f_min;
 
-	if (sdhc_clk / 16 > clock) {
-		for (; pre_div < 256; pre_div *= 2)
-			if ((sdhc_clk / pre_div) <= (clock *

16))

drivers/mmc/fsl_esdhc.c +1 -3
-				break;
-	} else
-		pre_div = 1;
+	while (sdhc_clk / (16 * pre_div * ddr_pre_div) >

clock && pre_div < 256)

drivers/mmc/fsl_esdhc.c +2 -3
+		pre_div *= 2;
 
-	for (div = 1; div <= 16; div++)
-		if ((sdhc_clk / (div * pre_div)) <= clock)
-			break;
+	while (sdhc_clk / (div * pre_div * ddr_pre_div) >

clock && div < 16)

drivers/mmc/fsl_esdhc.c +1 -0
+		div++;
Quoted 5 lines
Quoted 4 lines
drivers/mmc/fsl_esdhc.c +0 -1
 	pre_div >>= 1;
 	div -= 1;
--

2.11.0

Quoted 7 lines

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang

Denk

Quoted 3 lines

HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email:
lukma at denx.de

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang
Denk HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email:
lukma at denx.de

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email:
lukma at denx.de

#11 lore raw

PATCH 1/1 Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Quoted 142 lines

Subject: RE: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Quoted 107 lines

Subject: Re: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Quoted 8 lines · On Wed, 8 May 2019 13:59:14 +0000 Peng Fan <[email protected]> wrote:
Quoted 8 lines

-----Original Message-----
From: Lukasz Majewski [mailto:lukma at denx.de]
Sent: 2019年5月8日 14:38
To: Peng Fan <[email protected]>
Cc: u-boot at lists.denx.de; Tom Rini <[email protected]>; Marcel Ziswiler <[email protected]>; Fabio Estevam <[email protected]>; dl-uboot-imx <[email protected]>; sbabic at denx.de; Stefan Agner <[email protected]>; BOUGH

CHEN

Quoted 7 lines
Quoted 7 lines

<[email protected]>; Ye Li <[email protected]>
Subject: Re: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

On Wed, 8 May 2019 08:19:45 +0200
Lukasz Majewski <[email protected]> wrote:

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines

Hi Peng,

Quoted 6 lines

Hi Lukasz,

Quoted 4 lines

Subject: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

This reverts commit

72a89e0da5ac6a4ab929b15a2b656f04f50767f6,

Quoted 76 lines
Quoted 67 lines
Quoted 57 lines
Quoted 10 lines
Quoted 10 lines

which causes the imx53 HSC to hang as the eMMC is not working properly anymore.

The exact error message:
MMC write: dev # 0, block # 2, count 927 ... mmc write failed
0 blocks written: ERROR

imx53 is not using the DDR mode.

Quoted 8 lines
Quoted 6 lines

Debugging of pre_div and div generation showed that those values are generated in a way, which is not matching the ones from working setup.

As the original patch was performing code refactoring, let's revert this change, so all imx53 boards would work again.

Could you share what is the clock value for your board?

Sure, no problem:

Working setup:
--------------

MMC:
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240
FSL_SDHC: 0
Loading Environment from MMC...
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 1 div: 1 set_sysctl: clk: 272

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 1 div: 0 set_sysctl: clk: 256

Broken:
-------
MMC:
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240
FSL_SDHC: 0
Loading Environment from MMC...
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

Quoted 9 lines

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 0 div: 3 set_sysctl: clk: 48

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 0 div: 1 set_sysctl: clk: 16

(Please also find attached patch to reproduce debug output).

Quoted 6 lines

And maybe the most important question - why it was necessary to refactor this code?

Parts responsible for calculating pre_div and div seems not related to ddr problem (one spot issue is div <= 16 , but in the original code it was div < 16)?

Could you help verify whether the patch fixes you issue?

Quoted 12 lines

drivers/mmc/fsl_esdhc.c static void set_sysctl(struct fsl_esdhc_priv · line 643

Quoted 10 lines

index 1b7de74a72..3347fbe738 100644

drivers/mmc/fsl_esdhc.c +1 -2
@@ -640,8 +640,7 @@ static void set_sysctl(struct fsl_esdhc_priv
 *priv, struct mmc *mmc, uint clock) for (; pre_div < 256; pre_div *=
 2) if ((sdhc_clk / pre_div) <= (clock * 16))
                                break;
-       } else
-               pre_div = 1;
+       }

Please examine this code thoroughly and provide patch.

The pre_div should not override the initialization value at the beginning of the function.

Quoted 7 lines

The
} else

	pre_div = 1;

was added there for a purpose, so I'm wondering why it can be easily removed now.

The else was wrongly added. It is not correct.

Quoted 6 lines
Quoted 5 lines
drivers/mmc/fsl_esdhc.c +0 -0
        for (div = 1; div <= 16; div++)
                if ((sdhc_clk / (div * pre_div)) <= clock)
 
Quoted 6 lines
Quoted 3 lines

As I've stated above - is the above for() correct?

In the original code it was div < 16, but here it is div <= 16.

Checking i.MX53 SDHC DVS, it supports [1,16], so should use "<=16", other i.MX has same.

Should use "<16". Will pick up your revert to avoid regression.

Thanks,
Peng.

Quoted 128 lines

Divisor:
This register is used to provide a more exact divisor to generate the desired SD clock frequency.
NOTE: The divider can even support odd divisor without deterioration of duty cycle.
The setting are as following:
0x0 Divisor by 1
0x1 Divisor by 2
0xe Divisor by 15
0xf Divisor by 16

Regards,
Peng.

Quoted 71 lines
Quoted 71 lines

Thanks,
Peng.

Quoted 66 lines
Quoted 66 lines
Quoted 62 lines

Thanks,
Peng.

Quoted 56 lines
---
 drivers/mmc/fsl_esdhc.c | 23 +++++------------------
 1 file changed, 5 insertions(+), 18 deletions(-)
(quoted diff) +0 -0

b/drivers/mmc/fsl_esdhc.c index 1b7de74a72..377b2673a3 100644

drivers/mmc/fsl_esdhc.c +0 -0
@@ -621,31 +621,18 @@ static void set_sysctl(struct

fsl_esdhc_priv *priv, struct mmc *mmc, uint clock) #else

drivers/mmc/fsl_esdhc.c +1 -2
 	int pre_div = 2;
 #endif
+	int ddr_pre_div = mmc->ddr_mode ? 2 : 1;
 	int sdhc_clk = priv->sdhc_clk;
 	uint clk;
 
-	/*
-	 * For ddr mode, usdhc need to enable DDR mode first,

after select

drivers/mmc/fsl_esdhc.c +0 -1
-	 * this DDR mode, usdhc will automatically divide the

usdhc clock

drivers/mmc/fsl_esdhc.c +0 -3
-	 */
-	if (mmc->ddr_mode) {
-		writel(readl(&regs->mixctrl) |

MIX_CTRL_DDREN, &regs->mixctrl);

drivers/mmc/fsl_esdhc.c +0 -6
-		sdhc_clk >>= 1;
-	}
-
 	if (clock < mmc->cfg->f_min)
 		clock = mmc->cfg->f_min;
 
-	if (sdhc_clk / 16 > clock) {
-		for (; pre_div < 256; pre_div *= 2)
-			if ((sdhc_clk / pre_div) <= (clock *

16))

drivers/mmc/fsl_esdhc.c +1 -3
-				break;
-	} else
-		pre_div = 1;
+	while (sdhc_clk / (16 * pre_div * ddr_pre_div) >

clock && pre_div < 256)

drivers/mmc/fsl_esdhc.c +2 -3
+		pre_div *= 2;
 
-	for (div = 1; div <= 16; div++)
-		if ((sdhc_clk / (div * pre_div)) <= clock)
-			break;
+	while (sdhc_clk / (div * pre_div * ddr_pre_div) >

clock && div < 16)

drivers/mmc/fsl_esdhc.c +1 -1
+		div++;
 
 	pre_div >>= 1;
 	div -= 1;
--

2.11.0

Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director:

Wolfgang

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

Denk

Quoted 2 lines

HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80

Email:

Quoted 19 lines
Quoted 7 lines
Quoted 7 lines
Quoted 1 line

lukma at denx.de

Best regards,

Quoted 8 lines
Quoted 8 lines

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang
Denk HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email:
lukma at denx.de

Quoted 7 lines

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang

Denk

Quoted 3 lines

HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email:
lukma at denx.de

#12 lore raw

PATCH 1/1 Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Hi Lukasz,

Do you expect me to pick this patch, just see this patch was delegated to Stefano?

Regards,
Peng.

Quoted 289 lines

Subject: RE: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Quoted 143 lines

Subject: RE: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Quoted 107 lines

Subject: Re: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Quoted 8 lines · On Wed, 8 May 2019 13:59:14 +0000 Peng Fan <[email protected]> wrote:
Quoted 8 lines

-----Original Message-----
From: Lukasz Majewski [mailto:lukma at denx.de]
Sent: 2019年5月8日 14:38
To: Peng Fan <[email protected]>
Cc: u-boot at lists.denx.de; Tom Rini <[email protected]>; Marcel Ziswiler <[email protected]>; Fabio Estevam <[email protected]>; dl-uboot-imx <[email protected]>; sbabic at denx.de; Stefan Agner <[email protected]>; BOUGH

CHEN

Quoted 7 lines
Quoted 7 lines

<[email protected]>; Ye Li <[email protected]>
Subject: Re: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

On Wed, 8 May 2019 08:19:45 +0200 Lukasz Majewski
<[email protected]> wrote:

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines

Hi Peng,

Quoted 6 lines

Hi Lukasz,

Quoted 4 lines

Subject: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

This reverts commit

72a89e0da5ac6a4ab929b15a2b656f04f50767f6,

Quoted 76 lines
Quoted 67 lines
Quoted 57 lines
Quoted 10 lines
Quoted 10 lines

which causes the imx53 HSC to hang as the eMMC is not working properly anymore.

The exact error message:
MMC write: dev # 0, block # 2, count 927 ... mmc write failed
0 blocks written: ERROR

imx53 is not using the DDR mode.

Quoted 8 lines
Quoted 6 lines

Debugging of pre_div and div generation showed that those values are generated in a way, which is not matching the ones from working setup.

As the original patch was performing code refactoring, let's revert this change, so all imx53 boards would work again.

Could you share what is the clock value for your board?

Sure, no problem:

Working setup:
--------------

MMC:
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240
FSL_SDHC: 0
Loading Environment from MMC...
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 1 div: 1 set_sysctl: clk: 272

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 1 div: 0 set_sysctl: clk: 256

Broken:
-------
MMC:
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240
FSL_SDHC: 0
Loading Environment from MMC...
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

Quoted 9 lines

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 0 div: 3 set_sysctl: clk: 48

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 0 div: 1 set_sysctl: clk: 16

(Please also find attached patch to reproduce debug output).

Quoted 6 lines

And maybe the most important question - why it was necessary to refactor this code?

Parts responsible for calculating pre_div and div seems not related to ddr problem (one spot issue is div <= 16 , but in the original code it was div < 16)?

Could you help verify whether the patch fixes you issue?

Quoted 13 lines

drivers/mmc/fsl_esdhc.c static void set_sysctl(struct fsl_esdhc_priv · line 644

Quoted 11 lines

index 1b7de74a72..3347fbe738 100644

drivers/mmc/fsl_esdhc.c +1 -2
@@ -640,8 +640,7 @@ static void set_sysctl(struct fsl_esdhc_priv
 *priv, struct mmc *mmc, uint clock) for (; pre_div < 256; pre_div
 *=
 2) if ((sdhc_clk / pre_div) <= (clock * 16))
                                break;
-       } else
-               pre_div = 1;
+       }

Please examine this code thoroughly and provide patch.

The pre_div should not override the initialization value at the beginning of the function.

Quoted 7 lines

The
} else

	pre_div = 1;

was added there for a purpose, so I'm wondering why it can be easily removed now.

The else was wrongly added. It is not correct.

Quoted 6 lines
Quoted 5 lines
drivers/mmc/fsl_esdhc.c +0 -0
        for (div = 1; div <= 16; div++)
                if ((sdhc_clk / (div * pre_div)) <= clock)
Quoted 6 lines
Quoted 3 lines

As I've stated above - is the above for() correct?

In the original code it was div < 16, but here it is div <= 16.

Checking i.MX53 SDHC DVS, it supports [1,16], so should use "<=16", other i.MX has same.

Should use "<16". Will pick up your revert to avoid regression.

Thanks,
Peng.

Quoted 90 lines

Divisor:
This register is used to provide a more exact divisor to generate the desired SD clock frequency.
NOTE: The divider can even support odd divisor without deterioration of duty cycle.
The setting are as following:
0x0 Divisor by 1
0x1 Divisor by 2
0xe Divisor by 15
0xf Divisor by 16

Regards,
Peng.

Quoted 75 lines
Quoted 75 lines

Thanks,
Peng.

Quoted 70 lines
Quoted 70 lines
Quoted 57 lines

Thanks,
Peng.

Quoted 52 lines
---
 drivers/mmc/fsl_esdhc.c | 23 +++++------------------
 1 file changed, 5 insertions(+), 18 deletions(-)
(quoted diff) +0 -0

b/drivers/mmc/fsl_esdhc.c index 1b7de74a72..377b2673a3 100644

drivers/mmc/fsl_esdhc.c +0 -0
@@ -621,31 +621,18 @@ static void set_sysctl(struct

fsl_esdhc_priv *priv, struct mmc *mmc, uint clock) #else

drivers/mmc/fsl_esdhc.c +1 -2
 	int pre_div = 2;
 #endif
+	int ddr_pre_div = mmc->ddr_mode ? 2 : 1;
 	int sdhc_clk = priv->sdhc_clk;
 	uint clk;
 
-	/*
-	 * For ddr mode, usdhc need to enable DDR mode first,

after select

drivers/mmc/fsl_esdhc.c +0 -1
-	 * this DDR mode, usdhc will automatically divide the

usdhc clock

drivers/mmc/fsl_esdhc.c +0 -3
-	 */
-	if (mmc->ddr_mode) {
-		writel(readl(&regs->mixctrl) |

MIX_CTRL_DDREN, &regs->mixctrl);

drivers/mmc/fsl_esdhc.c +0 -6
-		sdhc_clk >>= 1;
-	}
-
 	if (clock < mmc->cfg->f_min)
 		clock = mmc->cfg->f_min;
 
-	if (sdhc_clk / 16 > clock) {
-		for (; pre_div < 256; pre_div *= 2)
-			if ((sdhc_clk / pre_div) <= (clock *

16))

drivers/mmc/fsl_esdhc.c +1 -3
-				break;
-	} else
-		pre_div = 1;
+	while (sdhc_clk / (16 * pre_div * ddr_pre_div) >

clock && pre_div < 256)

drivers/mmc/fsl_esdhc.c +2 -3
+		pre_div *= 2;
 
-	for (div = 1; div <= 16; div++)
-		if ((sdhc_clk / (div * pre_div)) <= clock)
-			break;
+	while (sdhc_clk / (div * pre_div * ddr_pre_div) >

clock && div < 16)

drivers/mmc/fsl_esdhc.c +1 -0
+		div++;
Quoted 5 lines
Quoted 4 lines
drivers/mmc/fsl_esdhc.c +0 -1
 	pre_div >>= 1;
 	div -= 1;
--

2.11.0

Best regards,

Lukasz Majewski

Quoted 20 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

--

DENX Software Engineering GmbH, Managing Director:

Wolfgang

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

Denk

Quoted 2 lines

HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80

Email:

Quoted 12 lines
Quoted 12 lines
Quoted 12 lines
Quoted 1 line

lukma at denx.de

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director:

Wolfgang

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

Denk HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-59 Fax:

(+49)-8142-66989-80 Email:

Quoted 17 lines
Quoted 5 lines
Quoted 1 line
Quoted 1 line

lukma at denx.de

Quoted 7 lines

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang

Denk

Quoted 4 lines

HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email:
lukma at denx.de

#13 lore raw

PATCH 1/1 Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Hi Peng,

Quoted 4 lines

Hi Lukasz,

Do you expect me to pick this patch, just see this patch was delegated to Stefano?

I think that you shall decide with Stefano who would take this patch.

For me there is no difference, I would just like to have it in main line ASAP (depends who prepare PR first).

Quoted 300 lines

Regards,
Peng.

Quoted 296 lines

Subject: RE: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Quoted 146 lines

Subject: RE: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Quoted 110 lines

Subject: Re: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Quoted 9 lines · On Wed, 8 May 2019 13:59:14 +0000 Peng Fan <[email protected]> wrote:
Quoted 9 lines

-----Original Message-----
From: Lukasz Majewski [mailto:lukma at denx.de]
Sent: 2019年5月8日 14:38
To: Peng Fan <[email protected]>
Cc: u-boot at lists.denx.de; Tom Rini <[email protected]>; Marcel Ziswiler <[email protected]>; Fabio Estevam <[email protected]>; dl-uboot-imx <[email protected]>; sbabic at denx.de; Stefan Agner <[email protected]>; BOUGH

CHEN

Quoted 7 lines
Quoted 7 lines

<[email protected]>; Ye Li <[email protected]>
Subject: Re: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

On Wed, 8 May 2019 08:19:45 +0200 Lukasz Majewski <[email protected]> wrote:

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines

Hi Peng,

Quoted 6 lines

Hi Lukasz,

Quoted 4 lines

Subject: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

This reverts commit

72a89e0da5ac6a4ab929b15a2b656f04f50767f6,

Quoted 78 lines
Quoted 69 lines
Quoted 62 lines
Quoted 14 lines
Quoted 14 lines

which causes the imx53 HSC to hang as the eMMC is not working properly anymore.

The exact error message:
MMC write: dev # 0, block # 2, count 927 ... mmc write failed
0 blocks written: ERROR

imx53 is not using the DDR mode.

Debugging of pre_div and div generation showed that those values are generated in a way, which is not matching the ones from working setup.

Quoted 6 lines
Quoted 3 lines

As the original patch was performing code refactoring, let's revert this change, so all imx53 boards would work again.

Could you share what is the clock value for your
board?

Sure, no problem:

Working setup:
--------------

MMC:
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240
FSL_SDHC: 0
Loading Environment from MMC...
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 1 div: 1 set_sysctl: clk: 272

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 1 div: 0 set_sysctl: clk: 256

Broken:
-------
MMC:
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240
FSL_SDHC: 0
Loading Environment from MMC...
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 0 div: 3 set_sysctl: clk: 48

Quoted 6 lines

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 0 div: 1 set_sysctl: clk: 16

(Please also find attached patch to reproduce debug output).

Quoted 6 lines

And maybe the most important question - why it was necessary to refactor this code?

Parts responsible for calculating pre_div and div seems not related to ddr problem (one spot issue is div <= 16 , but in the original code it was div < 16)?

Could you help verify whether the patch fixes you issue?

Quoted 13 lines

drivers/mmc/fsl_esdhc.c static void set_sysctl(struct fsl_esdhc_priv *priv, struct mmc *mmc, uint clock) · line 643

Quoted 11 lines

index 1b7de74a72..3347fbe738 100644

drivers/mmc/fsl_esdhc.c +0 -0
@@ -640,8 +640,7 @@ static void set_sysctl(struct
 fsl_esdhc_priv *priv, struct mmc *mmc, uint clock) for (;

pre_div < 256; pre_div *=
2) if ((sdhc_clk / pre_div) <= (clock * 16))

                                break;
drivers/mmc/fsl_esdhc.c +1 -2
-       } else
-               pre_div = 1;
+       }  

Please examine this code thoroughly and provide patch.

The pre_div should not override the initialization value at the beginning of the function.

Quoted 7 lines

The
} else

	pre_div = 1;

was added there for a purpose, so I'm wondering why it can be easily removed now.

The else was wrongly added. It is not correct.

Quoted 6 lines
Quoted 5 lines
 

        for (div = 1; div <= 16; div++)
                if ((sdhc_clk / (div * pre_div)) <= clock)
Quoted 7 lines
Quoted 4 lines

As I've stated above - is the above for() correct?

In the original code it was div < 16, but here it is div <= 16.

Checking i.MX53 SDHC DVS, it supports [1,16], so should use "<=16", other i.MX has same.

Should use "<16". Will pick up your revert to avoid regression.

Thanks,
Peng.

Quoted 93 lines

Divisor:
This register is used to provide a more exact divisor to generate the desired SD clock frequency.
NOTE: The divider can even support odd divisor without deterioration of duty cycle.
The setting are as following:
0x0 Divisor by 1
0x1 Divisor by 2
0xe Divisor by 15
0xf Divisor by 16

Regards,
Peng.

Quoted 78 lines
Quoted 78 lines

Thanks,
Peng.

Quoted 73 lines
Quoted 73 lines
Quoted 58 lines

Thanks,
Peng.

Quoted 53 lines
---
 drivers/mmc/fsl_esdhc.c | 23 +++++------------------
 1 file changed, 5 insertions(+), 18 deletions(-)
(quoted diff) +0 -0

b/drivers/mmc/fsl_esdhc.c index 1b7de74a72..377b2673a3 100644

drivers/mmc/fsl_esdhc.c +1 -2
@@ -621,31 +621,18 @@ static void set_sysctl(struct
 fsl_esdhc_priv *priv, struct mmc *mmc, uint clock)
 #else int pre_div = 2;
 #endif
+	int ddr_pre_div = mmc->ddr_mode ? 2 : 1;
 	int sdhc_clk = priv->sdhc_clk;
 	uint clk;
 
-	/*
-	 * For ddr mode, usdhc need to enable DDR

mode first, after select

drivers/mmc/fsl_esdhc.c +0 -1
-	 * this DDR mode, usdhc will automatically

divide the usdhc clock

drivers/mmc/fsl_esdhc.c +0 -3
-	 */
-	if (mmc->ddr_mode) {
-		writel(readl(&regs->mixctrl) |

MIX_CTRL_DDREN, &regs->mixctrl);

drivers/mmc/fsl_esdhc.c +0 -6
-		sdhc_clk >>= 1;
-	}
-
 	if (clock < mmc->cfg->f_min)
 		clock = mmc->cfg->f_min;
 
-	if (sdhc_clk / 16 > clock) {
-		for (; pre_div < 256; pre_div *= 2)
-			if ((sdhc_clk / pre_div) <=

(clock * 16))

drivers/mmc/fsl_esdhc.c +1 -3
-				break;
-	} else
-		pre_div = 1;
+	while (sdhc_clk / (16 * pre_div *

ddr_pre_div) > clock && pre_div < 256)

drivers/mmc/fsl_esdhc.c +1 -2
+		pre_div *= 2;
 
-	for (div = 1; div <= 16; div++)
-		if ((sdhc_clk / (div * pre_div)) <=

clock)

drivers/mmc/fsl_esdhc.c +1 -1
-			break;
+	while (sdhc_clk / (div * pre_div *

ddr_pre_div) > clock && div < 16)

drivers/mmc/fsl_esdhc.c +1 -0
+		div++;
Quoted 5 lines
Quoted 4 lines
drivers/mmc/fsl_esdhc.c +0 -1
 	pre_div >>= 1;
 	div -= 1;
--

2.11.0

Best regards,

Lukasz Majewski

--

Quoted 19 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

DENX Software Engineering GmbH, Managing Director:

Wolfgang

Quoted 4 lines
Quoted 4 lines
Quoted 4 lines

Denk

Quoted 3 lines

HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-59 Fax:
(+49)-8142-66989-80

Email:

Quoted 12 lines
Quoted 12 lines
Quoted 12 lines
Quoted 1 line

lukma at denx.de

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director:

Wolfgang

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

Denk HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-59 Fax:

(+49)-8142-66989-80 Email:

Quoted 18 lines
Quoted 5 lines
Quoted 1 line
Quoted 1 line

lukma at denx.de

Quoted 8 lines

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director:
Wolfgang

Denk

Quoted 4 lines

HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email:
lukma at denx.de

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: lukma at denx.de -------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 488 bytes
Desc: OpenPGP digital signature
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20190520/a7eed713/attachment-0001.sig>

#14 lore raw

PATCH 1/1 Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Hi Lukasz,

Quoted 7 lines

Subject: Re: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Hi Peng,

Quoted 2 lines

Hi Lukasz,

Quoted 7 lines
Quoted 2 lines

Do you expect me to pick this patch, just see this patch was delegated to Stefano?

I think that you shall decide with Stefano who would take this patch.

For me there is no difference, I would just like to have it in main line ASAP (depends who prepare PR first).

ok. I'll send a PR to Tom today after CI done. I was waiting IT to fix my ssh push issue, so delayed. I still have to use my github to PR (:

Regards,
Peng.

Quoted 312 lines
Quoted 298 lines

Regards,
Peng.

Quoted 293 lines

Subject: RE: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Quoted 145 lines

Subject: RE: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Quoted 109 lines

Subject: Re: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Quoted 9 lines · On Wed, 8 May 2019 13:59:14 +0000 Peng Fan <[email protected]> wrote:
Quoted 9 lines

-----Original Message-----
From: Lukasz Majewski [mailto:lukma at denx.de]
Sent: 2019年5月8日 14:38
To: Peng Fan <[email protected]>
Cc: u-boot at lists.denx.de; Tom Rini <[email protected]>; Marcel Ziswiler <[email protected]>; Fabio Estevam <[email protected]>; dl-uboot-imx <[email protected]>; sbabic at denx.de; Stefan Agner <[email protected]>; BOUGH

CHEN

Quoted 7 lines
Quoted 7 lines

<[email protected]>; Ye Li <[email protected]>
Subject: Re: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

On Wed, 8 May 2019 08:19:45 +0200 Lukasz Majewski <[email protected]> wrote:

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines

Hi Peng,

Quoted 6 lines

Hi Lukasz,

Quoted 4 lines

Subject: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

This reverts commit

72a89e0da5ac6a4ab929b15a2b656f04f50767f6,

Quoted 77 lines
Quoted 68 lines
Quoted 61 lines
Quoted 14 lines
Quoted 14 lines

which causes the imx53 HSC to hang as the eMMC is not working properly anymore.

The exact error message:
MMC write: dev # 0, block # 2, count 927 ... mmc write failed
0 blocks written: ERROR

imx53 is not using the DDR mode.

Debugging of pre_div and div generation showed that those values are generated in a way, which is not matching the ones from working setup.

Quoted 5 lines
Quoted 3 lines

As the original patch was performing code refactoring, let's revert this change, so all imx53 boards would work again.

Could you share what is the clock value for your board?

Sure, no problem:

Working setup:
--------------

MMC:
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240
FSL_SDHC: 0
Loading Environment from MMC...
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 1 div: 1 set_sysctl: clk: 272

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 1 div: 0 set_sysctl: clk: 256

Broken:
-------
MMC:
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240
FSL_SDHC: 0
Loading Environment from MMC...
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 0 div: 3 set_sysctl: clk: 48

Quoted 6 lines

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 0 div: 1 set_sysctl: clk: 16

(Please also find attached patch to reproduce debug output).

Quoted 6 lines

And maybe the most important question - why it was necessary to refactor this code?

Parts responsible for calculating pre_div and div seems not related to ddr problem (one spot issue is div <= 16 , but in the original code it was div < 16)?

Could you help verify whether the patch fixes you issue?

Quoted 13 lines

drivers/mmc/fsl_esdhc.c static void set_sysctl(struct fsl_esdhc_priv *priv, struct mmc *mmc, uint clock) · line 643

Quoted 11 lines

index 1b7de74a72..3347fbe738 100644

drivers/mmc/fsl_esdhc.c +0 -0
@@ -640,8 +640,7 @@ static void set_sysctl(struct
 fsl_esdhc_priv *priv, struct mmc *mmc, uint clock) for (;

pre_div < 256; pre_div *=
2) if ((sdhc_clk / pre_div) <= (clock * 16))

                                break;
drivers/mmc/fsl_esdhc.c +1 -2
-       } else
-               pre_div = 1;
+       }

Please examine this code thoroughly and provide patch.

The pre_div should not override the initialization value at the beginning of the function.

Quoted 7 lines

The
} else

	pre_div = 1;

was added there for a purpose, so I'm wondering why it can be easily removed now.

The else was wrongly added. It is not correct.

Quoted 6 lines
Quoted 5 lines

        for (div = 1; div <= 16; div++)
                if ((sdhc_clk / (div * pre_div)) <= clock)
Quoted 6 lines
Quoted 3 lines

As I've stated above - is the above for() correct?

In the original code it was div < 16, but here it is div <= 16.

Checking i.MX53 SDHC DVS, it supports [1,16], so should use "<=16", other i.MX has same.

Should use "<16". Will pick up your revert to avoid regression.

Thanks,
Peng.

Quoted 92 lines

Divisor:
This register is used to provide a more exact divisor to generate the desired SD clock frequency.
NOTE: The divider can even support odd divisor without deterioration of duty cycle.
The setting are as following:
0x0 Divisor by 1
0x1 Divisor by 2
0xe Divisor by 15
0xf Divisor by 16

Regards,
Peng.

Quoted 77 lines
Quoted 77 lines

Thanks,
Peng.

Quoted 72 lines
Quoted 72 lines
Quoted 57 lines

Thanks,
Peng.

Quoted 52 lines
---
 drivers/mmc/fsl_esdhc.c | 23 +++++------------------
 1 file changed, 5 insertions(+), 18 deletions(-)
(quoted diff) +0 -0

b/drivers/mmc/fsl_esdhc.c index 1b7de74a72..377b2673a3 100644

drivers/mmc/fsl_esdhc.c +1 -2
@@ -621,31 +621,18 @@ static void set_sysctl(struct
 fsl_esdhc_priv *priv, struct mmc *mmc, uint clock)
 #else int pre_div = 2;  #endif
+	int ddr_pre_div = mmc->ddr_mode ? 2 : 1;
 	int sdhc_clk = priv->sdhc_clk;
 	uint clk;
 
-	/*
-	 * For ddr mode, usdhc need to enable DDR

mode first, after select

drivers/mmc/fsl_esdhc.c +0 -1
-	 * this DDR mode, usdhc will automatically

divide the usdhc clock

drivers/mmc/fsl_esdhc.c +0 -3
-	 */
-	if (mmc->ddr_mode) {
-		writel(readl(&regs->mixctrl) |

MIX_CTRL_DDREN, &regs->mixctrl);

drivers/mmc/fsl_esdhc.c +0 -6
-		sdhc_clk >>= 1;
-	}
-
 	if (clock < mmc->cfg->f_min)
 		clock = mmc->cfg->f_min;
 
-	if (sdhc_clk / 16 > clock) {
-		for (; pre_div < 256; pre_div *= 2)
-			if ((sdhc_clk / pre_div) <=

(clock * 16))

drivers/mmc/fsl_esdhc.c +1 -3
-				break;
-	} else
-		pre_div = 1;
+	while (sdhc_clk / (16 * pre_div *

ddr_pre_div) > clock && pre_div < 256)

drivers/mmc/fsl_esdhc.c +1 -2
+		pre_div *= 2;
 
-	for (div = 1; div <= 16; div++)
-		if ((sdhc_clk / (div * pre_div)) <=

clock)

drivers/mmc/fsl_esdhc.c +1 -1
-			break;
+	while (sdhc_clk / (div * pre_div *

ddr_pre_div) > clock && div < 16)

drivers/mmc/fsl_esdhc.c +1 -0
+		div++;
Quoted 5 lines
Quoted 4 lines
drivers/mmc/fsl_esdhc.c +0 -1
 	pre_div >>= 1;
 	div -= 1;
--

2.11.0

Best regards,

Lukasz Majewski

--

Quoted 19 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

DENX Software Engineering GmbH, Managing Director:

Wolfgang

Quoted 4 lines
Quoted 4 lines
Quoted 4 lines

Denk

Quoted 3 lines

HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-59 Fax:
(+49)-8142-66989-80

Email:

Quoted 12 lines
Quoted 12 lines
Quoted 12 lines
Quoted 1 line

lukma at denx.de

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director:

Wolfgang

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

Denk HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-59 Fax:

(+49)-8142-66989-80 Email:

Quoted 18 lines
Quoted 5 lines
Quoted 1 line
Quoted 1 line

lukma at denx.de

Quoted 8 lines

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director:
Wolfgang

Denk

Quoted 4 lines

HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email:
lukma at denx.de

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email:
lukma at denx.de

#15 lore raw

PATCH 1/1 Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Hi Peng,

Quoted 12 lines

Hi Lukasz,

Quoted 10 lines

Subject: Re: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Hi Peng,

Quoted 4 lines

Hi Lukasz,

Do you expect me to pick this patch, just see this patch was delegated to Stefano?

Quoted 7 lines
Quoted 5 lines

I think that you shall decide with Stefano who would take this patch.

For me there is no difference, I would just like to have it in main line ASAP (depends who prepare PR first).

ok. I'll send a PR to Tom today after CI done.

Great. Thanks.

Quoted 3 lines

I was waiting IT to
fix my ssh push issue, so delayed. I still have to use my github to PR (:

I do know this topic well ... (Large companies have very strict rules regarding perceiving firewalls).

Quoted 323 lines

Regards,
Peng.

Quoted 319 lines
Quoted 305 lines

Regards,
Peng.

Quoted 300 lines

Subject: RE: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Quoted 146 lines

Subject: RE: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Quoted 110 lines

Subject: Re: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

Quoted 9 lines · On Wed, 8 May 2019 13:59:14 +0000 Peng Fan <[email protected]> wrote:
Quoted 9 lines

-----Original Message-----
From: Lukasz Majewski [mailto:lukma at denx.de]
Sent: 2019年5月8日 14:38
To: Peng Fan <[email protected]>
Cc: u-boot at lists.denx.de; Tom Rini <[email protected]>; Marcel Ziswiler <[email protected]>; Fabio Estevam <[email protected]>; dl-uboot-imx <[email protected]>; sbabic at denx.de; Stefan Agner <[email protected]>; BOUGH

CHEN

Quoted 7 lines
Quoted 7 lines

<[email protected]>; Ye Li <[email protected]>
Subject: Re: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

On Wed, 8 May 2019 08:19:45 +0200 Lukasz Majewski <[email protected]> wrote:

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines

Hi Peng,

Quoted 6 lines

Hi Lukasz,

Quoted 4 lines

Subject: [PATCH] Revert "mmc: fsl_esdhc: fix sd/mmc ddr mode clock setting issue"

This reverts commit

72a89e0da5ac6a4ab929b15a2b656f04f50767f6,

Quoted 78 lines
Quoted 69 lines
Quoted 62 lines
Quoted 14 lines
Quoted 14 lines

which causes the imx53 HSC to hang as the eMMC is not working properly anymore.

The exact error message:
MMC write: dev # 0, block # 2, count 927 ... mmc write failed
0 blocks written: ERROR

imx53 is not using the DDR mode.

Debugging of pre_div and div generation showed that those values are generated in a way, which is not matching the ones from working setup.

Quoted 6 lines
Quoted 3 lines

As the original patch was performing code refactoring, let's revert this change, so all imx53 boards would work again.

Could you share what is the clock value for your
board?

Sure, no problem:

Working setup:
--------------

MMC:
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240
FSL_SDHC: 0
Loading Environment from MMC...
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 1 div: 1 set_sysctl: clk: 272

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 1 div: 0 set_sysctl: clk: 256

Broken:
-------
MMC:
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240
FSL_SDHC: 0
Loading Environment from MMC...
set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 8 div: 12 set_sysctl: clk: 2240

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 0 div: 3 set_sysctl: clk: 48

Quoted 6 lines

set_sysctl: pre_div = 2 mmc->ddr: 0 sdhc_clk: 80000000
pre_div: 0 div: 1 set_sysctl: clk: 16

(Please also find attached patch to reproduce debug output).

Quoted 6 lines

And maybe the most important question - why it was necessary to refactor this code?

Parts responsible for calculating pre_div and div seems not related to ddr problem (one spot issue is div <= 16 , but in the original code it was div < 16)?

Could you help verify whether the patch fixes you issue?

Quoted 13 lines

drivers/mmc/fsl_esdhc.c

Quoted 11 lines

index 1b7de74a72..3347fbe738 100644

drivers/mmc/fsl_esdhc.c +0 -0
@@ -640,8 +640,7 @@ static void set_sysctl(struct
 fsl_esdhc_priv *priv, struct mmc *mmc, uint clock) for (;

pre_div < 256; pre_div *=
2) if ((sdhc_clk / pre_div) <= (clock * 16))

                                break;
drivers/mmc/fsl_esdhc.c +1 -2
-       } else
-               pre_div = 1;
+       }  

Please examine this code thoroughly and provide patch.

The pre_div should not override the initialization value at the beginning of the function.

Quoted 7 lines

The
} else

	pre_div = 1;

was added there for a purpose, so I'm wondering why it can be easily removed now.

The else was wrongly added. It is not correct.

Quoted 6 lines
Quoted 5 lines
 

        for (div = 1; div <= 16; div++)
                if ((sdhc_clk / (div * pre_div)) <= clock)
Quoted 7 lines
Quoted 4 lines

As I've stated above - is the above for() correct?

In the original code it was div < 16, but here it is div <= 16.

Checking i.MX53 SDHC DVS, it supports [1,16], so should use "<=16", other i.MX has same.

Should use "<16". Will pick up your revert to avoid regression.

Thanks,
Peng.

Quoted 95 lines

Divisor:
This register is used to provide a more exact divisor to generate the desired SD clock frequency.
NOTE: The divider can even support odd divisor without deterioration of duty cycle.
The setting are as following:
0x0 Divisor by 1
0x1 Divisor by 2
0xe Divisor by 15
0xf Divisor by 16

Regards,
Peng.

Quoted 80 lines
Quoted 80 lines

Thanks,
Peng.

Quoted 75 lines
Quoted 75 lines
Quoted 60 lines

Thanks,
Peng.

Quoted 55 lines
---
 drivers/mmc/fsl_esdhc.c | 23

+++++------------------ 1 file changed, 5 insertions(+), 18 deletions(-)

(quoted diff) +0 -0

b/drivers/mmc/fsl_esdhc.c index
1b7de74a72..377b2673a3 100644

drivers/mmc/fsl_esdhc.c +0 -0
@@ -621,31 +621,18 @@ static void

set_sysctl(struct fsl_esdhc_priv *priv, struct mmc *mmc, uint clock) #else int pre_div = 2; #endif

drivers/mmc/fsl_esdhc.c +1 -2
+	int ddr_pre_div = mmc->ddr_mode ? 2 : 1;
 	int sdhc_clk = priv->sdhc_clk;
 	uint clk;
 
-	/*
-	 * For ddr mode, usdhc need to enable DDR

mode first, after select

drivers/mmc/fsl_esdhc.c +0 -1
-	 * this DDR mode, usdhc will

automatically divide the usdhc clock

drivers/mmc/fsl_esdhc.c +0 -3
-	 */
-	if (mmc->ddr_mode) {
-		writel(readl(&regs->mixctrl) |

MIX_CTRL_DDREN, &regs->mixctrl);

drivers/mmc/fsl_esdhc.c +0 -5
-		sdhc_clk >>= 1;
-	}
-
 	if (clock < mmc->cfg->f_min)
 		clock = mmc->cfg->f_min;
 
-	if (sdhc_clk / 16 > clock) {
-		for (; pre_div < 256; pre_div *=

2)

drivers/mmc/fsl_esdhc.c +0 -1
-			if ((sdhc_clk / pre_div)

<= (clock * 16))

drivers/mmc/fsl_esdhc.c +1 -3
-				break;
-	} else
-		pre_div = 1;
+	while (sdhc_clk / (16 * pre_div *

ddr_pre_div) > clock && pre_div < 256)

drivers/mmc/fsl_esdhc.c +1 -2
+		pre_div *= 2;
 
-	for (div = 1; div <= 16; div++)
-		if ((sdhc_clk / (div * pre_div))

<= clock)

drivers/mmc/fsl_esdhc.c +1 -1
-			break;
+	while (sdhc_clk / (div * pre_div *

ddr_pre_div) > clock && div < 16)

drivers/mmc/fsl_esdhc.c +1 -0
+		div++;
Quoted 5 lines
Quoted 4 lines
drivers/mmc/fsl_esdhc.c +0 -1
 	pre_div >>= 1;
 	div -= 1;
--

2.11.0

Best regards,

Lukasz Majewski

--

Quoted 21 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

DENX Software Engineering GmbH, Managing
Director:

Wolfgang

Quoted 4 lines
Quoted 4 lines
Quoted 4 lines

Denk

Quoted 3 lines

HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-59 Fax:
(+49)-8142-66989-80

Email:

Quoted 13 lines
Quoted 13 lines
Quoted 13 lines
Quoted 1 line

lukma at denx.de

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing
Director:

Wolfgang

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

Denk HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-59 Fax:

(+49)-8142-66989-80 Email:

Quoted 18 lines
Quoted 5 lines
Quoted 1 line
Quoted 1 line

lukma at denx.de

Quoted 8 lines

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director:
Wolfgang

Denk

Quoted 4 lines

HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email:
lukma at denx.de

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang
Denk HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email:
lukma at denx.de

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: lukma at denx.de -------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 488 bytes
Desc: OpenPGP digital signature
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20190520/7de80f41/attachment.sig>