[U-Boot] [PATCH] powerpc/t1040qds: Add DDR Raw Timing support

[U-Boot] [PATCH] powerpc/t1040qds: Add DDR Raw Timing support

[Priyanka Jain]

T1040QDS-D3 has dual-rank DDR:
	Micron, MT18KSF51272AZ-1G6 (4GB, x72, CL=10).
Add Raw Timing structure for this DDR.

Typically SPD method is used for getting DDR parameter
and calculating values for various DDR controller registers.

But somentimes it may happen that SPD present on DDR may get
accidently erased or is not working properly during initial
bring-up. In that circumnstance, DDR raw timing structure can be
use as fallback option for getting DDR parameters.

Signed-off-by: Poonam Aggrwal <poonam.aggrwal@freescale.com>
Signed-off-by: Priyanka Jain <Priyanka.Jain@freescale.com>
---
 Based on u-boot-mpc85xx/next branch.
 board/freescale/t1040qds/ddr.c |   15 +++++++++++++++
 board/freescale/t1040qds/ddr.h |   29 +++++++++++++++++++++++++++++
 include/configs/T1040QDS.h     |    1 +
 3 files changed, 45 insertions(+), 0 deletions(-)

diff --git a/board/freescale/t1040qds/ddr.c b/board/freescale/t1040qds/ddr.c
index 4fd17da..16ab829 100644
--- a/board/freescale/t1040qds/ddr.c
+++ b/board/freescale/t1040qds/ddr.c
@@ -15,6 +15,21 @@
 
 DECLARE_GLOBAL_DATA_PTR;
 
+int fsl_ddr_get_dimm_params(dimm_params_t *pdimm,
+		unsigned int controller_number,
+		unsigned int dimm_number)
+{
+	const char dimm_model[] = "RAW timing DDR";
+
+	if ((controller_number == 0) && (dimm_number == 0)) {
+		memcpy(pdimm, &ddr_raw_timing, sizeof(dimm_params_t));
+		memset(pdimm->mpart, 0, sizeof(pdimm->mpart));
+		memcpy(pdimm->mpart, dimm_model, sizeof(dimm_model) - 1);
+	}
+
+	return 0;
+}
+
 void fsl_ddr_board_options(memctl_options_t *popts,
 				dimm_params_t *pdimm,
 				unsigned int ctrl_num)
diff --git a/board/freescale/t1040qds/ddr.h b/board/freescale/t1040qds/ddr.h
index 8ee206e..4a4f76a 100644
--- a/board/freescale/t1040qds/ddr.h
+++ b/board/freescale/t1040qds/ddr.h
@@ -6,6 +6,35 @@
 
 #ifndef __DDR_H__
 #define __DDR_H__
+dimm_params_t ddr_raw_timing = {
+	.n_ranks = 2,
+	.rank_density = 2147483648u,
+	.capacity = 4294967296u,
+	.primary_sdram_width = 64,
+	.ec_sdram_width = 8,
+	.registered_dimm = 0,
+	.mirrored_dimm = 0,
+	.n_row_addr = 15,
+	.n_col_addr = 10,
+	.n_banks_per_sdram_device = 8,
+	.edc_config = 2,	/* ECC */
+	.burst_lengths_bitmask = 0x0c,
+	.tckmin_x_ps = 1071,
+	.caslat_x = 0xfe << 4,	/* 5,6,7,8,9,10,11 */
+	.taa_ps = 13125,
+	.twr_ps = 15000,
+	.trcd_ps = 13125,
+	.trrd_ps = 6000,
+	.trp_ps = 13125,
+	.tras_ps = 34000,
+	.trc_ps = 48125,
+	.trfc_ps = 260000,
+	.twtr_ps = 7500,
+	.trtp_ps = 7500,
+	.refresh_rate_ps = 7800000,
+	.tfaw_ps = 35000,
+};
+
 struct board_specific_parameters {
 	u32 n_ranks;
 	u32 datarate_mhz_high;
diff --git a/include/configs/T1040QDS.h b/include/configs/T1040QDS.h
index 2738242..a8b32fe 100644
--- a/include/configs/T1040QDS.h
+++ b/include/configs/T1040QDS.h
@@ -168,6 +168,7 @@ unsigned long get_board_ddr_clk(void);
 #define CONFIG_CHIP_SELECTS_PER_CTRL	(4 * CONFIG_DIMM_SLOTS_PER_CTLR)
 
 #define CONFIG_DDR_SPD
+#define CONFIG_SYS_DDR_RAW_TIMING
 #define CONFIG_FSL_DDR3
 #define CONFIG_FSL_DDR_INTERACTIVE
 
-- 
1.7.4.1

[U-Boot] [PATCH] powerpc/t1040qds: Add DDR Raw Timing support

[York Sun]

On 10/25/2013 02:49 AM, Priyanka Jain wrote:
> T1040QDS-D3 has dual-rank DDR:
> 	Micron, MT18KSF51272AZ-1G6 (4GB, x72, CL=10).
> Add Raw Timing structure for this DDR.
> 
> Typically SPD method is used for getting DDR parameter
> and calculating values for various DDR controller registers.
> 
> But somentimes it may happen that SPD present on DDR may get
> accidently erased or is not working properly during initial
> bring-up. In that circumnstance, DDR raw timing structure can be
> use as fallback option for getting DDR parameters.
> 
> Signed-off-by: Poonam Aggrwal <poonam.aggrwal@freescale.com>
> Signed-off-by: Priyanka Jain <Priyanka.Jain@freescale.com>
> ---

I don't think it is appropriate to use raw timing for this application.
This board use DIMMs, not soldered DDR chips. In normal cases, this
piece of code is not used. It is only useful if you have a DIMM without
a valid SPD. You could get another DIMM. I am not saying this patch is
wrong, it is just useless for general public.

York

[U-Boot] [PATCH] powerpc/t1040qds: Add DDR Raw Timing support

[Jain Priyanka-B32167]

> -----Original Message-----
> From: sun york-R58495
> Sent: Wednesday, October 30, 2013 10:14 PM
> To: Jain Priyanka-B32167; u-boot at lists.denx.de
> Cc: Aggrwal Poonam-B10812
> Subject: Re: [PATCH] powerpc/t1040qds: Add DDR Raw Timing support
> 
> On 10/25/2013 02:49 AM, Priyanka Jain wrote:
> > T1040QDS-D3 has dual-rank DDR:
> > 	Micron, MT18KSF51272AZ-1G6 (4GB, x72, CL=10).
> > Add Raw Timing structure for this DDR.
> >
> > Typically SPD method is used for getting DDR parameter and calculating
> > values for various DDR controller registers.
> >
> > But somentimes it may happen that SPD present on DDR may get
> > accidently erased or is not working properly during initial bring-up.
> > In that circumnstance, DDR raw timing structure can be use as fallback
> > option for getting DDR parameters.
> >
> > Signed-off-by: Poonam Aggrwal <poonam.aggrwal@freescale.com>
> > Signed-off-by: Priyanka Jain <Priyanka.Jain@freescale.com>
> > ---
> 
> I don't think it is appropriate to use raw timing for this application.
> This board use DIMMs, not soldered DDR chips. In normal cases, this piece
> of code is not used. It is only useful if you have a DIMM without a valid
> SPD. You could get another DIMM. I am not saying this patch is wrong, it
> is just useless for general public.
> 
> York

[Jain Priyanka-B32167]  You are right that in case of DIMM type of memories, raw timing may not be required. 
But there are certain scenarios in which keeping raw timing is useful like
1)During initial bring-up, there is a possibility that I2C bus may be broken, 
In that case boot-loader has to rely on the fallback option of raw timing.
2)Also raw timings are used in emulator to test t1040.
So, we prefer to keep raw timing as fallback option.

Regards
Priyanka

[U-Boot] [PATCH] powerpc/t1040qds: Add DDR Raw Timing support

[sun york-R58495]

On Oct 30, 2013, at 9:55 PM, Jain Priyanka-B32167 wrote:

> 
>> -----Original Message-----
>> From: sun york-R58495
>> Sent: Wednesday, October 30, 2013 10:14 PM
>> To: Jain Priyanka-B32167; u-boot at lists.denx.de
>> Cc: Aggrwal Poonam-B10812
>> Subject: Re: [PATCH] powerpc/t1040qds: Add DDR Raw Timing support
>> 
>> On 10/25/2013 02:49 AM, Priyanka Jain wrote:
>>> T1040QDS-D3 has dual-rank DDR:
>>> 	Micron, MT18KSF51272AZ-1G6 (4GB, x72, CL=10).
>>> Add Raw Timing structure for this DDR.
>>> 
>>> Typically SPD method is used for getting DDR parameter and calculating
>>> values for various DDR controller registers.
>>> 
>>> But somentimes it may happen that SPD present on DDR may get
>>> accidently erased or is not working properly during initial bring-up.
>>> In that circumnstance, DDR raw timing structure can be use as fallback
>>> option for getting DDR parameters.
>>> 
>>> Signed-off-by: Poonam Aggrwal <poonam.aggrwal@freescale.com>
>>> Signed-off-by: Priyanka Jain <Priyanka.Jain@freescale.com>
>>> ---
>> 
>> I don't think it is appropriate to use raw timing for this application.
>> This board use DIMMs, not soldered DDR chips. In normal cases, this piece
>> of code is not used. It is only useful if you have a DIMM without a valid
>> SPD. You could get another DIMM. I am not saying this patch is wrong, it
>> is just useless for general public.
>> 
>> York
> 
> [Jain Priyanka-B32167]  You are right that in case of DIMM type of memories, raw timing may not be required. 
> But there are certain scenarios in which keeping raw timing is useful like
> 1)During initial bring-up, there is a possibility that I2C bus may be broken, 
> In that case boot-loader has to rely on the fallback option of raw timing.
> 2)Also raw timings are used in emulator to test t1040.
> So, we prefer to keep raw timing as fallback option.

It is a debug technique, not an implementation. You can use fixed register setting as well for debugging. But it doesn't mean you have to keep it in the final code. After you cleaning up the code, all "fixed" code should be removed, including the raw timing parameters you are using. Emulator target also has I2C EEPROM modeled. You should use SPD as well.

York

[U-Boot] [PATCH] powerpc/t1040qds: Add DDR Raw Timing support

[Priyanka Jain]

Ok,
I will drop this patch and maintain it locally.

Regards
Priyanka


> -----Original Message-----
> From: sun york-R58495
> Sent: Thursday, October 31, 2013 10:50 AM
> To: Jain Priyanka-B32167
> Cc: sun york-R58495; u-boot at lists.denx.de; Aggrwal Poonam-B10812
> Subject: Re: [PATCH] powerpc/t1040qds: Add DDR Raw Timing support
> 
> 
> On Oct 30, 2013, at 9:55 PM, Jain Priyanka-B32167 wrote:
> 
> >
> >> -----Original Message-----
> >> From: sun york-R58495
> >> Sent: Wednesday, October 30, 2013 10:14 PM
> >> To: Jain Priyanka-B32167; u-boot at lists.denx.de
> >> Cc: Aggrwal Poonam-B10812
> >> Subject: Re: [PATCH] powerpc/t1040qds: Add DDR Raw Timing support
> >>
> >> On 10/25/2013 02:49 AM, Priyanka Jain wrote:
> >>> T1040QDS-D3 has dual-rank DDR:
> >>> 	Micron, MT18KSF51272AZ-1G6 (4GB, x72, CL=10).
> >>> Add Raw Timing structure for this DDR.
> >>>
> >>> Typically SPD method is used for getting DDR parameter and
> >>> calculating values for various DDR controller registers.
> >>>
> >>> But somentimes it may happen that SPD present on DDR may get
> >>> accidently erased or is not working properly during initial bring-up.
> >>> In that circumnstance, DDR raw timing structure can be use as
> >>> fallback option for getting DDR parameters.
> >>>
> >>> Signed-off-by: Poonam Aggrwal <poonam.aggrwal@freescale.com>
> >>> Signed-off-by: Priyanka Jain <Priyanka.Jain@freescale.com>
> >>> ---
> >>
> >> I don't think it is appropriate to use raw timing for this
> application.
> >> This board use DIMMs, not soldered DDR chips. In normal cases, this
> >> piece of code is not used. It is only useful if you have a DIMM
> >> without a valid SPD. You could get another DIMM. I am not saying this
> >> patch is wrong, it is just useless for general public.
> >>
> >> York
> >
> > [Jain Priyanka-B32167]  You are right that in case of DIMM type of
> memories, raw timing may not be required.
> > But there are certain scenarios in which keeping raw timing is useful
> > like 1)During initial bring-up, there is a possibility that I2C bus
> > may be broken, In that case boot-loader has to rely on the fallback
> option of raw timing.
> > 2)Also raw timings are used in emulator to test t1040.
> > So, we prefer to keep raw timing as fallback option.
> 
> It is a debug technique, not an implementation. You can use fixed
> register setting as well for debugging. But it doesn't mean you have to
> keep it in the final code. After you cleaning up the code, all "fixed"
> code should be removed, including the raw timing parameters you are
> using. Emulator target also has I2C EEPROM modeled. You should use SPD as
> well.
> 
> York