iMX8MPlus + DDR4 stuck at wait_ddrphy_training_complete()

[khang_letruong]

Dear Community,

We are bringing up a custom iMX8MP based board using DDR4 (Micron MT40A512M16LY). We were able to follow the steps described in MSCALE_DDR_Tool_User_Guide.pdf and the DDR Calibration completed with success.

However, when integrating the generated ddr4_timing.c into u-Boot lf_v2021.04 aligned to BSP-5.10.52_2.1.0, the board got stuck at the while loop within wait_ddrphy_training_complete()

U-Boot SPL 2021.04 (Jan 12 2023 - 16:20:19 +0700)
DDRINFO: start DRAM init
DDRINFO: cfg clk
DDRINFO: DRAM rate 3200MTS
DDRINFO: ddrc config start
DDRINFO: ddrc config done
DDRINFO:ddrphy config start
DRAM PHY training for 3200MTS
check ddr_pmu_train_imem code
check ddr_pmu_train_imem code pass
check ddr4_pmu_train_dmem code
check ddr_pmu_train_dmem code pass
Enter wait_ddrphy_training_complete() // I added this at the begining of wait_ddrphy_training_complete()

Could you help to point out what could be the issue, please?

In attachment is our custom RPA.

Thanks in advance and best regards,

Khang

[Rita_Wang]

Could you share the DDR datasheet and your DDR section schematic to us?

[khang_letruong]

Hi @Rita_Wang ,

I shared with you the DDR4 part of schematic as well as the datasheet of the DDR4 in private message.

Regards,
K.

[khang_letruong]

Hi @Rita_Wang ,

I was able to go a little bit further : not being stuck anymore but training reported FAILED for rate = 1336MTS :

What could be the issue, please ?

Best Regards,
K

[khang_letruong]

Hi again @Rita_Wang ,

Precisely, our DDR4 is MT40A512M16LY-062:E and it has data rate 3200 in the below table :

So naturally the training is passed at that speed but not others. Do you think so? If yes, how to "by-pass" other training speeds, please?

It looks like there was a similar issue supported by @igorpadykov about iMX8MM with similar DDR4 family.

Regards,
K.

[khang_letruong]

Hi,

Relate to training rate 1336, I found that it appears in the end of our generated ddr4_timing.c :

struct dram_fsp_msg ddr_dram_fsp_msg[] = {
    {
        /* P0 3200mts 1D */
        .drate = 3200,
        .fw_type = FW_1D_IMAGE,
        .fsp_cfg = ddr_fsp0_cfg,
        .fsp_cfg_num = ARRAY_SIZE(ddr_fsp0_cfg),
    },
    {
        /* P1 1336mts 1D */
        .drate = 1336,
        .fw_type = FW_1D_IMAGE,
        .fsp_cfg = ddr_fsp1_cfg,
        .fsp_cfg_num = ARRAY_SIZE(ddr_fsp1_cfg),
    },
    {
        /* P0 3200mts 2D */
        .drate = 3200,
        .fw_type = FW_2D_IMAGE,
        .fsp_cfg = ddr_fsp0_2d_cfg,
        .fsp_cfg_num = ARRAY_SIZE(ddr_fsp0_2d_cfg),
    },
};

/* ddr timing config params */
struct dram_timing_info dram_timing = {
    .ddrc_cfg = ddr_ddrc_cfg,
    .ddrc_cfg_num = ARRAY_SIZE(ddr_ddrc_cfg),
    .ddrphy_cfg = ddr_ddrphy_cfg,
    .ddrphy_cfg_num = ARRAY_SIZE(ddr_ddrphy_cfg),
    .fsp_msg = ddr_dram_fsp_msg,
    .fsp_msg_num = ARRAY_SIZE(ddr_dram_fsp_msg),
    .ddrphy_trained_csr = ddr_ddrphy_trained_csr,
    .ddrphy_trained_csr_num = ARRAY_SIZE(ddr_ddrphy_trained_csr),
    .ddrphy_pie = ddr_phy_pie,
    .ddrphy_pie_num = ARRAY_SIZE(ddr_phy_pie),
    .fsp_table = { 3200, 1336, },
};

Meanwhile, in the ddr4_timing.c of iMX8MP-EVK, 2nd data rate is 1066 :

struct dram_fsp_msg ddr_dram_fsp_msg[] = {
{
/* P0 3200mts 1D */
.drate = 3200,
.fw_type = FW_1D_IMAGE,
.fsp_cfg = ddr_fsp0_cfg,
.fsp_cfg_num = ARRAY_SIZE(ddr_fsp0_cfg),
},
{
/* P1 1066mts 1D */
.drate = 1066,
.fw_type = FW_1D_IMAGE,
.fsp_cfg = ddr_fsp1_cfg,
.fsp_cfg_num = ARRAY_SIZE(ddr_fsp1_cfg),
},
{
/* P0 3200mts 2D */
.drate = 3200,
.fw_type = FW_2D_IMAGE,
.fsp_cfg = ddr_fsp0_2d_cfg,
.fsp_cfg_num = ARRAY_SIZE(ddr_fsp0_2d_cfg),
},
};

/* ddr timing config params */
struct dram_timing_info dram_timing = {
.ddrc_cfg = ddr_ddrc_cfg,
.ddrc_cfg_num = ARRAY_SIZE(ddr_ddrc_cfg),
.ddrphy_cfg = ddr_ddrphy_cfg,
.ddrphy_cfg_num = ARRAY_SIZE(ddr_ddrphy_cfg),
.fsp_msg = ddr_dram_fsp_msg,
.fsp_msg_num = ARRAY_SIZE(ddr_dram_fsp_msg),
.ddrphy_trained_csr = ddr_ddrphy_trained_csr,
.ddrphy_trained_csr_num = ARRAY_SIZE(ddr_ddrphy_trained_csr),
.ddrphy_pie = ddr_phy_pie,
.ddrphy_pie_num = ARRAY_SIZE(ddr_phy_pie),
.fsp_table = { 3200, 1066, },
};

I tried to modify the RPA by changing number of frequency setpoints from 2 to 1 :

But MSCALE_DDR_Tool denied to run the stress test :

I then changed the Clock to specific value of my DDR4 (1600) instead of Default: 

but the ddr4_timing.c is the same as before (i.e 1336 still appeared).

Not sure how to remove that 1336 data rate out of the ddr4_timing.c.

Regards,
K

[khang_letruong · Accepted solution]

HI @Rita_Wang ,

I resolved the issue by being able to select 1 for Number of frequency setpoints 

If I kept Number of frequency setpoints as 2, I think it would also work of I modified FREQ1 setpoints Clock Cycle Freq (MHz) from 668 to 533 for having data rate 1066 as the case of other references (including iMX8MP-EVK with DDR4) but I haven't tested yet.

Thanks for your support.

K

[Rita_Wang]

Good news it works.

[khang_letruong]

Hi @Rita_Wang ,

Thank you. But now I am facing another problem : https://community.nxp.com/t5/i-MX-Processors/iMX8MP-DDR4-bringup-SPL-failed-to-boot-from-all-boot-de...

Could you also provide support, please ?

Best Regards,

Khang