Skip to contentSkip to conversation

arm: imx: mmdc_size: Increase row_max for imx8m

PATCH Applied Teresa Remmet 3 participants 8 messages May 20, 2022 – Jun 21, 2022

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 arm: imx: mmdc_size: Increase row_max for imx8m

As DDRC_ADDRMAP7_ROW_B16 and DDRC_ADDRMAP7_ROW_B17 are used now for the row size calculation we need to increase row_max to 18.

For LPDDR4 this only works in combination with ram timings created with recent configuration spreadsheet versions. With older versions the register DDRC_ADDRMAP7 may not be set and calculation will lead to wrong results even with this patch.

Fixes dad2b5636bd8 ("ARM: imx: Add imx8 support for 18 bit SDRAM row size handle")
---
 arch/arm/mach-imx/esdctl.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/arm/mach-imx/esdctl.c +1 -1
index 8dd0ddbbc965..b070ebc62a45 100644
@@ -488,7 +488,7 @@ static resource_size_t imx8m_ddrc_sdram_size(void __iomem *ddrc)
 
 	return imx_ddrc_sdram_size(ddrc, addrmap,
 				   12, ARRAY_AND_SIZE(col_b),
-				   16, ARRAY_AND_SIZE(row_b),
+				   18, ARRAY_AND_SIZE(row_b),
 				   reduced_adress_space, true);
 }
 
Signature · 7 lines
2.25.1


_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
Applied
#1 lore raw
Quoted 8 lines · Teresa Remmet

As DDRC_ADDRMAP7_ROW_B16 and DDRC_ADDRMAP7_ROW_B17 are used now for the row size calculation we need to increase row_max to 18.

For LPDDR4 this only works in combination with ram timings created with recent configuration spreadsheet versions. With older versions the register DDRC_ADDRMAP7 may not be set and calculation will lead to wrong results even with this patch.

Quoted 5 lines
Fixes dad2b5636bd8 ("ARM: imx: Add imx8 support for 18 bit SDRAM row size handle")
---
 arch/arm/mach-imx/esdctl.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Applied, thanks

Sascha

Quoted 23 lines
arch/arm/mach-imx/esdctl.c +1 -1
index 8dd0ddbbc965..b070ebc62a45 100644
@@ -488,7 +488,7 @@ static resource_size_t imx8m_ddrc_sdram_size(void __iomem *ddrc)
 
 	return imx_ddrc_sdram_size(ddrc, addrmap,
 				   12, ARRAY_AND_SIZE(col_b),
-				   16, ARRAY_AND_SIZE(row_b),
+				   18, ARRAY_AND_SIZE(row_b),
 				   reduced_adress_space, true);
 }
 
Signature · 7 lines
2.25.1


_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
Signature · 9 lines
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
#2 lore raw
In reply to Teresa Remmet

Hello,

Quoted 10 lines · Teresa Remmet

As DDRC_ADDRMAP7_ROW_B16 and DDRC_ADDRMAP7_ROW_B17 are used now for the row size calculation we need to increase row_max to 18.

For LPDDR4 this only works in combination with ram timings created with recent configuration spreadsheet versions. With older versions the register DDRC_ADDRMAP7 may not be set and calculation will lead to wrong results even with this patch.

Fixes dad2b5636bd8 ("ARM: imx: Add imx8 support for 18 bit SDRAM row size handle")

I have an out-of-tree i.MX8MM board with LPDDR4 that reported a correct size of 1G prior to this patch and now it reports 4G. DDRC_ADDRMAP7 is not explicitly initialized, but it's being read as zero, so the calculation seems to still be broken..

Quoted 17 lines
---
 arch/arm/mach-imx/esdctl.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/arm/mach-imx/esdctl.c +1 -1
index 8dd0ddbbc965..b070ebc62a45 100644
@@ -488,7 +488,7 @@ static resource_size_t imx8m_ddrc_sdram_size(void __iomem *ddrc)
 
 	return imx_ddrc_sdram_size(ddrc, addrmap,
 				   12, ARRAY_AND_SIZE(col_b),
-				   16, ARRAY_AND_SIZE(row_b),
+				   18, ARRAY_AND_SIZE(row_b),
 				   reduced_adress_space, true);
 }
 
Signature · 4 lines
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |
#3 lore raw

Hello Ahmad,

Am Montag, dem 20.06.2022 um 14:02 +0200 schrieb Ahmad Fatoum:

Quoted 15 lines

Hello,

Quoted 11 lines · Teresa Remmet

As DDRC_ADDRMAP7_ROW_B16 and DDRC_ADDRMAP7_ROW_B17 are used now for the row size calculation we need to increase row_max to 18.

For LPDDR4 this only works in combination with ram timings created with recent configuration spreadsheet versions. With older versions the register DDRC_ADDRMAP7 may not be set and calculation will lead to wrong results even with this patch.

Fixes dad2b5636bd8 ("ARM: imx: Add imx8 support for 18 bit SDRAM

row size handle")

Quoted 6 lines

I have an out-of-tree i.MX8MM board with LPDDR4 that reported a correct size of 1G
prior to this patch and now it reports 4G. DDRC_ADDRMAP7 is not explicitly initialized,
but it's being read as zero, so the calculation seems to still be broken..

yes, this is why I added the note to the commit message. The RAM timings of the board you are using have been created with a old version of the spreadsheet ( < version 19 for i.MX8MM).
This issue is worked around when this patch is reverted as DDRC_ADDRMAP7 is not taken into account. But calculating big ram sizes will probably not work then.

I have set the DDRC_ADDRMAP7 register manually in the RAM configuration in such a case. As I don't see a solution that fits for all. But would be happy for one. :)

Regards,
Teresa

Quoted 22 lines
Quoted 20 lines
---
 arch/arm/mach-imx/esdctl.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/arm/mach- renamed from arch/arm/mach-imx/esdctl.c +0 -0

imx/esdctl.c
index 8dd0ddbbc965..b070ebc62a45 100644

arch/arm/mach-imx/esdctl.c +1 -1
@@ -488,7 +488,7 @@ static resource_size_t
 imx8m_ddrc_sdram_size(void __iomem *ddrc)
 
 	return imx_ddrc_sdram_size(ddrc, addrmap,
 				   12, ARRAY_AND_SIZE(col_b),
-				   16, ARRAY_AND_SIZE(row_b),
+				   18, ARRAY_AND_SIZE(row_b),
 				   reduced_adress_space, true);
 }
Signature · 5 lines
PHYTEC Messtechnik GmbH | Robert-Koch-Str. 39 | 55129 Mainz, Germany

Geschäftsführer: Dipl.-Ing. Michael Mitezki, Dipl.-Ing. Bodo Huber |
Handelsregister Mainz HRB 4656 | Finanzamt Mainz | St.Nr. 266500608, DE
149059855
#4 lore raw

Hello Teresa,

Quoted 32 lines · Teresa Remmet

Hello Ahmad,

Am Montag, dem 20.06.2022 um 14:02 +0200 schrieb Ahmad Fatoum:

Quoted 15 lines

Hello,

Quoted 11 lines · Teresa Remmet

As DDRC_ADDRMAP7_ROW_B16 and DDRC_ADDRMAP7_ROW_B17 are used now for the row size calculation we need to increase row_max to 18.

For LPDDR4 this only works in combination with ram timings created with recent configuration spreadsheet versions. With older versions the register DDRC_ADDRMAP7 may not be set and calculation will lead to wrong results even with this patch.

Fixes dad2b5636bd8 ("ARM: imx: Add imx8 support for 18 bit SDRAM

row size handle")

Quoted 6 lines

I have an out-of-tree i.MX8MM board with LPDDR4 that reported a correct size of 1G
prior to this patch and now it reports 4G. DDRC_ADDRMAP7 is not explicitly initialized,
but it's being read as zero, so the calculation seems to still be broken..

yes, this is why I added the note to the commit message. The RAM timings of the board you are using have been created with a old version of the spreadsheet ( < version 19 for i.MX8MM).
This issue is worked around when this patch is reverted as DDRC_ADDRMAP7 is not taken into account. But calculating big ram sizes will probably not work then.

Quoted 3 lines

I have set the DDRC_ADDRMAP7 register manually in the RAM configuration in such a case. As I don't see a solution that fits for all. But would be happy for one. :)

What would the 'neutral' value to write into this register be? zero seems to not be it.

Thanks,
Ahmad

Quoted 26 lines

Regards,
Teresa

Quoted 22 lines
Quoted 20 lines
---
 arch/arm/mach-imx/esdctl.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/arm/mach- renamed from arch/arm/mach-imx/esdctl.c +0 -0

imx/esdctl.c
index 8dd0ddbbc965..b070ebc62a45 100644

arch/arm/mach-imx/esdctl.c +1 -1
@@ -488,7 +488,7 @@ static resource_size_t
 imx8m_ddrc_sdram_size(void __iomem *ddrc)
 
 	return imx_ddrc_sdram_size(ddrc, addrmap,
 				   12, ARRAY_AND_SIZE(col_b),
-				   16, ARRAY_AND_SIZE(row_b),
+				   18, ARRAY_AND_SIZE(row_b),
 				   reduced_adress_space, true);
 }
Signature · 4 lines
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |
#5 lore raw

Am Montag, dem 20.06.2022 um 14:38 +0200 schrieb Ahmad Fatoum:

Quoted 40 lines

Hello Teresa,

Quoted 37 lines · Teresa Remmet

Hello Ahmad,

Am Montag, dem 20.06.2022 um 14:02 +0200 schrieb Ahmad Fatoum:

Quoted 18 lines

Hello,

Quoted 10 lines · Teresa Remmet

As DDRC_ADDRMAP7_ROW_B16 and DDRC_ADDRMAP7_ROW_B17 are used now for the row size calculation we need to increase row_max to 18.

For LPDDR4 this only works in combination with ram timings created with recent configuration spreadsheet versions. With older versions the register DDRC_ADDRMAP7 may not be set and
calculation will lead to wrong results even with this patch.

Quoted 4 lines
Fixes dad2b5636bd8 ("ARM: imx: Add imx8 support for 18 bit

SDRAM
row size handle")

Quoted 6 lines

I have an out-of-tree i.MX8MM board with LPDDR4 that reported a correct size of 1G
prior to this patch and now it reports 4G. DDRC_ADDRMAP7 is not explicitly initialized,
but it's being read as zero, so the calculation seems to still be broken..

yes, this is why I added the note to the commit message. The RAM timings of the board you are using have been created with a old version
of the spreadsheet ( < version 19 for i.MX8MM).
This issue is worked around when this patch is reverted as DDRC_ADDRMAP7 is not taken into account. But calculating big ram sizes
will probably not work then.

Quoted 8 lines
Quoted 5 lines

I have set the DDRC_ADDRMAP7 register manually in the RAM configuration
in such a case. As I don't see a solution that fits for all. But would
be happy for one. :)

What would the 'neutral' value to write into this register be? zero seems to not be it.

it's

0xf0f for ADDRMAP7.

Reference Manual says: "If set to 15, row address bit X is set to 0"

Regards,
Teresa

Quoted 28 lines

Thanks,
Ahmad

Quoted 22 lines

Regards,
Teresa

Quoted 19 lines
Quoted 19 lines
---
 arch/arm/mach-imx/esdctl.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/arm/mach- renamed from arch/arm/mach-imx/esdctl.c +0 -0

imx/esdctl.c
index 8dd0ddbbc965..b070ebc62a45 100644

arch/arm/mach-imx/esdctl.c +1 -1
@@ -488,7 +488,7 @@ static resource_size_t
 imx8m_ddrc_sdram_size(void __iomem *ddrc)
 
 	return imx_ddrc_sdram_size(ddrc, addrmap,
 				   12, ARRAY_AND_SIZE(col_b),
-				   16, ARRAY_AND_SIZE(row_b),
+				   18, ARRAY_AND_SIZE(row_b),
 				   reduced_adress_space, true);
 }
Signature · 5 lines
PHYTEC Messtechnik GmbH | Robert-Koch-Str. 39 | 55129 Mainz, Germany

Geschäftsführer: Dipl.-Ing. Michael Mitezki, Dipl.-Ing. Bodo Huber |
Handelsregister Mainz HRB 4656 | Finanzamt Mainz | St.Nr. 266500608, DE
149059855
#6 lore raw

Hi,

Quoted 8 lines · Teresa Remmet

Am Montag, dem 20.06.2022 um 14:38 +0200 schrieb Ahmad Fatoum:

Quoted 7 lines · Teresa Remmet
Quoted 5 lines

I have set the DDRC_ADDRMAP7 register manually in the RAM configuration
in such a case. As I don't see a solution that fits for all. But would
be happy for one. :)

Quoted 8 lines
Quoted 2 lines

What would the 'neutral' value to write into this register be? zero seems to not be it.

it's

0xf0f for ADDRMAP7.

Reference Manual says: "If set to 15, row address bit X is set to 0"

Do newer spreadsheets always generate ADDRMAP7 writes even if the value is zero? If so, we could perhaps initialize it to 0xf0f before running ddr_cfg_umctl2(). The DDRC seems to be in reset while the registers are being written, so this might just work.

As 0 is a valid value, I am wondering if this snippet introduced with 42d45ef380c5 ("ARM: imx: Add imx8 support for SDRAM with two or more bank groups") is correct:

if (addrmap[8]) {
        if (FIELD_GET(DDRC_ADDRMAP8_BG_B0, addrmap[8]) != 0b11111)
                banks++;
        if (FIELD_GET(DDRC_ADDRMAP8_BG_B1, addrmap[8]) != 0b111111)
                banks++;
}

Thanks,
Ahmad

Quoted 33 lines

Regards,
Teresa

Quoted 28 lines

Thanks,
Ahmad

Quoted 22 lines

Regards,
Teresa

Quoted 19 lines
Quoted 19 lines
---
 arch/arm/mach-imx/esdctl.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/arm/mach- renamed from arch/arm/mach-imx/esdctl.c +0 -0

imx/esdctl.c
index 8dd0ddbbc965..b070ebc62a45 100644

arch/arm/mach-imx/esdctl.c +1 -1
@@ -488,7 +488,7 @@ static resource_size_t
 imx8m_ddrc_sdram_size(void __iomem *ddrc)
 
 	return imx_ddrc_sdram_size(ddrc, addrmap,
 				   12, ARRAY_AND_SIZE(col_b),
-				   16, ARRAY_AND_SIZE(row_b),
+				   18, ARRAY_AND_SIZE(row_b),
 				   reduced_adress_space, true);
 }
Signature · 4 lines
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |
#7 lore raw

Hi!

Am Montag, dem 20.06.2022 um 15:19 +0200 schrieb Ahmad Fatoum:

Quoted 20 lines

Hi,

Quoted 17 lines · Teresa Remmet

Am Montag, dem 20.06.2022 um 14:38 +0200 schrieb Ahmad Fatoum:

Quoted 11 lines · Teresa Remmet
Quoted 6 lines

I have set the DDRC_ADDRMAP7 register manually in the RAM configuration
in such a case. As I don't see a solution that fits for all. But
would
be happy for one. :)

What would the 'neutral' value to write into this register be? zero
seems to not be it.

it's

0xf0f for ADDRMAP7.

Quoted 7 lines
Quoted 2 lines

Reference Manual says: "If set to 15, row address bit X is set to 0"

Do newer spreadsheets always generate ADDRMAP7 writes even if the value is zero? If so, we could perhaps initialize it to 0xf0f before running ddr_cfg_umctl2(). The DDRC seems to be in reset while the registers are being written, so this might just work.

this is not so easy to figure out because the spread sheet does not generate the c headers directly.
But it is true for other ADDRMAP registers where the reset value is also zero. So I would give it a try.

Quoted 6 lines

As 0 is a valid value, I am wondering if this snippet introduced with 42d45ef380c5 ("ARM: imx: Add imx8 support for SDRAM with two or more bank groups")
is correct:

Quoted 6 lines
if (addrmap[8]) {
        if (FIELD_GET(DDRC_ADDRMAP8_BG_B0, addrmap[8]) != 0b11111)
                banks++;
        if (FIELD_GET(DDRC_ADDRMAP8_BG_B1, addrmap[8]) != 0b111111)
                banks++;
}

yes, this is wrong. LPDDR4 is no problem at the moment as the spreadsheet does not set the value and the reset value here is zero. But DDR4 uses it and there could be a "real" zero set which is then ignored.

Regards,
Teresa

Quoted 35 lines

Thanks,
Ahmad

Quoted 29 lines

Regards,
Teresa

Quoted 25 lines

Thanks,
Ahmad

Quoted 22 lines

Regards,
Teresa

Quoted 19 lines
Quoted 19 lines
---
 arch/arm/mach-imx/esdctl.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/arm/mach- renamed from arch/arm/mach-imx/esdctl.c +0 -0

imx/esdctl.c
index 8dd0ddbbc965..b070ebc62a45 100644

arch/arm/mach-imx/esdctl.c +1 -1
@@ -488,7 +488,7 @@ static resource_size_t
 imx8m_ddrc_sdram_size(void __iomem *ddrc)
 
 	return imx_ddrc_sdram_size(ddrc, addrmap,
 				   12, ARRAY_AND_SIZE(col_b),
-				   16, ARRAY_AND_SIZE(row_b),
+				   18, ARRAY_AND_SIZE(row_b),
 				   reduced_adress_space, true);
 }
Signature · 5 lines
PHYTEC Messtechnik GmbH | Robert-Koch-Str. 39 | 55129 Mainz, Germany

Geschäftsführer: Dipl.-Ing. Michael Mitezki, Dipl.-Ing. Bodo Huber |
Handelsregister Mainz HRB 4656 | Finanzamt Mainz | St.Nr. 266500608, DE
149059855