From: Andre Przywara <andre.przywara@arm.com>
To: Jernej Skrabec <jernej.skrabec@gmail.com>
Cc: jagan@amarulasolutions.com, u-boot@lists.denx.de
Subject: Re: [PATCH 7/8] sunxi: Parameterize bit delay code in H616 DRAM driver
Date: Wed, 4 Jan 2023 00:37:47 +0000 [thread overview]
Message-ID: <20230104003747.2d5a16f6@slackpad.lan> (raw)
In-Reply-To: <20221211163213.98540-8-jernej.skrabec@gmail.com>
On Sun, 11 Dec 2022 17:32:12 +0100
Jernej Skrabec <jernej.skrabec@gmail.com> wrote:
Hi Jernej,
> These values are highly board specific and thus make sense to add
> parameter for them. To ease adding support for new boards, let's make
> them same as in vendor DRAM settings.
So scrolling up and down: does this patch miss the TPR11 and TPR12
values in the OPi-Zero2 defconfig? And should we not default to 0 in
Kconfig to help spotting this omission more easily for new boards?
If I pieced the bits together correctly, we end up with the same values
in the register with TPR11=0xfffedddb and TPR12=0xeddca998, and ODT_EN
being irrelevant.
> Signed-off-by: Jernej Skrabec <jernej.skrabec@gmail.com>
> ---
> .../include/asm/arch-sunxi/dram_sun50i_h616.h | 4 +
> arch/arm/mach-sunxi/Kconfig | 18 ++
> arch/arm/mach-sunxi/dram_sun50i_h616.c | 189 +++++++++++++-----
> 3 files changed, 162 insertions(+), 49 deletions(-)
>
> diff --git a/arch/arm/include/asm/arch-sunxi/dram_sun50i_h616.h b/arch/arm/include/asm/arch-sunxi/dram_sun50i_h616.h
> index b5140c79b70e..c7890c83391f 100644
> --- a/arch/arm/include/asm/arch-sunxi/dram_sun50i_h616.h
> +++ b/arch/arm/include/asm/arch-sunxi/dram_sun50i_h616.h
> @@ -145,6 +145,7 @@ check_member(sunxi_mctl_ctl_reg, unk_0x4240, 0x4240);
> #define TPR10_READ_CALIBRATION BIT(21)
> #define TPR10_READ_TRAINING BIT(22)
> #define TPR10_WRITE_TRAINING BIT(23)
> +#define TPR10_UNKNOWN_FEAT3 BIT(30)
As mentioned in the other patch: if we don't know the meaning of this
bit, I'd prefer using BIT(30) directly, or at least encode BIT30
in the name.
>
> struct dram_para {
> u32 clk;
> @@ -156,7 +157,10 @@ struct dram_para {
> u32 dx_odt;
> u32 dx_dri;
> u32 ca_dri;
> + u32 odt_en;
> u32 tpr10;
> + u32 tpr11;
> + u32 tpr12;
> };
>
>
> diff --git a/arch/arm/mach-sunxi/Kconfig b/arch/arm/mach-sunxi/Kconfig
> index 778304b77e26..b050f0a56971 100644
> --- a/arch/arm/mach-sunxi/Kconfig
> +++ b/arch/arm/mach-sunxi/Kconfig
> @@ -67,11 +67,29 @@ config DRAM_SUN50I_H616_CA_DRI
> help
> CA DRI value from vendor DRAM settings.
>
> +config DRAM_SUN50I_H616_ODT_EN
> + hex "H616 DRAM ODT EN parameter"
> + default 0x1
> + help
> + ODT EN value from vendor DRAM settings.
> +
> config DRAM_SUN50I_H616_TPR10
> hex "H616 DRAM TPR10 parameter"
> help
> TPR10 value from vendor DRAM settings. It tells which features
> should be configured, like write leveling, read calibration, etc.
> +
> +config DRAM_SUN50I_H616_TPR11
> + hex "H616 DRAM TPR11 parameter"
> + default 0x0
> + help
> + TPR11 value from vendor DRAM settings.
> +
> +config DRAM_SUN50I_H616_TPR12
> + hex "H616 DRAM TPR12 parameter"
> + default 0x0
> + help
> + TPR12 value from vendor DRAM settings.
> endif
>
> config SUN6I_PRCM
> diff --git a/arch/arm/mach-sunxi/dram_sun50i_h616.c b/arch/arm/mach-sunxi/dram_sun50i_h616.c
> index 3b2ba168498c..df06cea42464 100644
> --- a/arch/arm/mach-sunxi/dram_sun50i_h616.c
> +++ b/arch/arm/mach-sunxi/dram_sun50i_h616.c
> @@ -574,7 +574,7 @@ static bool mctl_phy_write_training(struct dram_para *para)
>
> static void mctl_phy_bit_delay_compensation(struct dram_para *para)
> {
> - u32 *ptr;
> + u32 *ptr, val;
> int i;
>
> if (para->tpr10 & TPR10_UNKNOWN_FEAT2) {
> @@ -582,49 +582,93 @@ static void mctl_phy_bit_delay_compensation(struct dram_para *para)
> setbits_le32(SUNXI_DRAM_PHY0_BASE + 8, 8);
> clrbits_le32(SUNXI_DRAM_PHY0_BASE + 0x190, 0x10);
>
> + if (para->tpr10 & TPR10_UNKNOWN_FEAT3)
> + val = para->tpr11 & 0x3f;
> + else
> + val = (para->tpr11 & 0xf) << 1;
> +
> ptr = (u32 *)(SUNXI_DRAM_PHY0_BASE + 0x484);
> for (i = 0; i < 9; i++) {
> - writel_relaxed(0x16, ptr);
> - writel_relaxed(0x16, ptr + 0x30);
> + writel_relaxed(val, ptr);
> + writel_relaxed(val, ptr + 0x30);
> ptr += 2;
> }
> - writel_relaxed(0x1c, SUNXI_DRAM_PHY0_BASE + 0x4d0);
> - writel_relaxed(0x1c, SUNXI_DRAM_PHY0_BASE + 0x590);
> - writel_relaxed(0x1c, SUNXI_DRAM_PHY0_BASE + 0x4cc);
> - writel_relaxed(0x1c, SUNXI_DRAM_PHY0_BASE + 0x58c);
> +
> + if (para->tpr10 & TPR10_UNKNOWN_FEAT3)
> + val = (para->odt_en >> 15) & 0x1e;
So I guess odt_en stands for "ODT enable". Looking at the D1 DRAM
driver, they have a boot0 parameter odt_en which is either 0x0 or 0x1,
so it looks like a boolean value. This seems to be also the case here?
In the D1 driver, this seems to gate the ZQ value being used, which
provides the actual timing values.
So is TPR10_UNKNOWN_FEAT3 actually this ODT_EN switch, and the variable
containing the timing bits should be zq_value or CONFIG_DRAM_ZQ?
> + else
> + val = (para->tpr11 >> 15) & 0x1e;
So I am wondering if it's less confusing to mask first, then shift?
Granted, this here is shift from [19:16] into [5:1], so it's not easy
either way, but I think it's easier to match the mask with the provided
TPR11 value.
> +
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x4d0);
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x590);
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x4cc);
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x58c);
> +
> + if (para->tpr10 & TPR10_UNKNOWN_FEAT3)
> + val = (para->tpr11 >> 8) & 0x3f;
> + else
> + val = (para->tpr11 >> 3) & 0x1e;
>
> ptr = (u32 *)(SUNXI_DRAM_PHY0_BASE + 0x4d8);
> for (i = 0; i < 9; i++) {
> - writel_relaxed(0x1a, ptr);
> - writel_relaxed(0x1a, ptr + 0x30);
> + writel_relaxed(val, ptr);
> + writel_relaxed(val, ptr + 0x30);
> ptr += 2;
> }
> - writel_relaxed(0x1e, SUNXI_DRAM_PHY0_BASE + 0x524);
> - writel_relaxed(0x1e, SUNXI_DRAM_PHY0_BASE + 0x5e4);
> - writel_relaxed(0x1e, SUNXI_DRAM_PHY0_BASE + 0x520);
> - writel_relaxed(0x1e, SUNXI_DRAM_PHY0_BASE + 0x5e0);
> +
> + if (para->tpr10 & TPR10_UNKNOWN_FEAT3)
> + val = (para->odt_en >> 19) & 0x1e;
Is this really odt_en as a base, while *most* of the other bits use
TPR11 as well? Just asking, since it doesn't fit into my naive pattern
matching ;-)
The rest looks correct (minus the repeat of the above mentioned
issues), I compared the register offsets between - and +.
Cheers,
Andre
> + else
> + val = (para->tpr11 >> 19) & 0x1e;
> +
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x524);
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x5e4);
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x520);
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x5e0);
> +
> + if (para->tpr10 & TPR10_UNKNOWN_FEAT3)
> + val = (para->tpr11 >> 16) & 0x3f;
> + else
> + val = (para->tpr11 >> 7) & 0x1e;
>
> ptr = (u32 *)(SUNXI_DRAM_PHY0_BASE + 0x604);
> for (i = 0; i < 9; i++) {
> - writel_relaxed(0x1a, ptr);
> - writel_relaxed(0x1a, ptr + 0x30);
> + writel_relaxed(val, ptr);
> + writel_relaxed(val, ptr + 0x30);
> ptr += 2;
> }
> - writel_relaxed(0x1e, SUNXI_DRAM_PHY0_BASE + 0x650);
> - writel_relaxed(0x1e, SUNXI_DRAM_PHY0_BASE + 0x710);
> - writel_relaxed(0x1e, SUNXI_DRAM_PHY0_BASE + 0x64c);
> - writel_relaxed(0x1e, SUNXI_DRAM_PHY0_BASE + 0x70c);
> +
> + if (para->tpr10 & TPR10_UNKNOWN_FEAT3)
> + val = (para->odt_en >> 23) & 0x1e;
> + else
> + val = (para->tpr11 >> 23) & 0x1e;
> +
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x650);
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x710);
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x64c);
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x70c);
> +
> + if (para->tpr10 & TPR10_UNKNOWN_FEAT3)
> + val = (para->tpr11 >> 24) & 0x3f;
> + else
> + val = (para->tpr11 >> 11) & 0x1e;
>
> ptr = (u32 *)(SUNXI_DRAM_PHY0_BASE + 0x658);
> for (i = 0; i < 9; i++) {
> - writel_relaxed(0x1a, ptr);
> - writel_relaxed(0x1a, ptr + 0x30);
> + writel_relaxed(val, ptr);
> + writel_relaxed(val, ptr + 0x30);
> ptr += 2;
> }
> - writel_relaxed(0x1e, SUNXI_DRAM_PHY0_BASE + 0x6a4);
> - writel_relaxed(0x1e, SUNXI_DRAM_PHY0_BASE + 0x764);
> - writel_relaxed(0x1e, SUNXI_DRAM_PHY0_BASE + 0x6a0);
> - writel_relaxed(0x1e, SUNXI_DRAM_PHY0_BASE + 0x760);
> +
> + if (para->tpr10 & TPR10_UNKNOWN_FEAT3)
> + val = (para->odt_en >> 27) & 0x1e;
> + else
> + val = (para->tpr11 >> 27) & 0x1e;
> +
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x6a4);
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x764);
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x6a0);
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x760);
>
> dmb();
>
> @@ -635,49 +679,93 @@ static void mctl_phy_bit_delay_compensation(struct dram_para *para)
> clrbits_le32(SUNXI_DRAM_PHY0_BASE + 0x54, 0x80);
> clrbits_le32(SUNXI_DRAM_PHY0_BASE + 0x190, 4);
>
> + if (para->tpr10 & TPR10_UNKNOWN_FEAT3)
> + val = para->tpr12 & 0x3f;
> + else
> + val = (para->tpr12 & 0xf) << 1;
> +
> ptr = (u32 *)(SUNXI_DRAM_PHY0_BASE + 0x480);
> for (i = 0; i < 9; i++) {
> - writel_relaxed(0x10, ptr);
> - writel_relaxed(0x10, ptr + 0x30);
> + writel_relaxed(val, ptr);
> + writel_relaxed(val, ptr + 0x30);
> ptr += 2;
> }
> - writel_relaxed(0x18, SUNXI_DRAM_PHY0_BASE + 0x528);
> - writel_relaxed(0x18, SUNXI_DRAM_PHY0_BASE + 0x5e8);
> - writel_relaxed(0x18, SUNXI_DRAM_PHY0_BASE + 0x4c8);
> - writel_relaxed(0x18, SUNXI_DRAM_PHY0_BASE + 0x588);
> +
> + if (para->tpr10 & TPR10_UNKNOWN_FEAT3)
> + val = (para->odt_en << 1) & 0x1e;
> + else
> + val = (para->tpr12 >> 15) & 0x1e;
> +
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x528);
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x5e8);
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x4c8);
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x588);
> +
> + if (para->tpr10 & TPR10_UNKNOWN_FEAT3)
> + val = (para->tpr12 >> 8) & 0x3f;
> + else
> + val = (para->tpr12 >> 3) & 0x1e;
>
> ptr = (u32 *)(SUNXI_DRAM_PHY0_BASE + 0x4d4);
> for (i = 0; i < 9; i++) {
> - writel_relaxed(0x12, ptr);
> - writel_relaxed(0x12, ptr + 0x30);
> + writel_relaxed(val, ptr);
> + writel_relaxed(val, ptr + 0x30);
> ptr += 2;
> }
> - writel_relaxed(0x1a, SUNXI_DRAM_PHY0_BASE + 0x52c);
> - writel_relaxed(0x1a, SUNXI_DRAM_PHY0_BASE + 0x5ec);
> - writel_relaxed(0x1a, SUNXI_DRAM_PHY0_BASE + 0x51c);
> - writel_relaxed(0x1a, SUNXI_DRAM_PHY0_BASE + 0x5dc);
> +
> + if (para->tpr10 & TPR10_UNKNOWN_FEAT3)
> + val = (para->odt_en >> 3) & 0x1e;
> + else
> + val = (para->tpr12 >> 19) & 0x1e;
> +
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x52c);
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x5ec);
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x51c);
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x5dc);
> +
> + if (para->tpr10 & TPR10_UNKNOWN_FEAT3)
> + val = (para->tpr12 >> 16) & 0x3f;
> + else
> + val = (para->tpr12 >> 7) & 0x1e;
>
> ptr = (u32 *)(SUNXI_DRAM_PHY0_BASE + 0x600);
> for (i = 0; i < 9; i++) {
> - writel_relaxed(0x12, ptr);
> - writel_relaxed(0x12, ptr + 0x30);
> + writel_relaxed(val, ptr);
> + writel_relaxed(val, ptr + 0x30);
> ptr += 2;
> }
> - writel_relaxed(0x1a, SUNXI_DRAM_PHY0_BASE + 0x6a8);
> - writel_relaxed(0x1a, SUNXI_DRAM_PHY0_BASE + 0x768);
> - writel_relaxed(0x1a, SUNXI_DRAM_PHY0_BASE + 0x648);
> - writel_relaxed(0x1a, SUNXI_DRAM_PHY0_BASE + 0x708);
> +
> + if (para->tpr10 & TPR10_UNKNOWN_FEAT3)
> + val = (para->odt_en >> 7) & 0x1e;
> + else
> + val = (para->tpr12 >> 23) & 0x1e;
> +
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x6a8);
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x768);
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x648);
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x708);
> +
> + if (para->tpr10 & TPR10_UNKNOWN_FEAT3)
> + val = (para->tpr12 >> 24) & 0x3f;
> + else
> + val = (para->tpr12 >> 11) & 0x1e;
>
> ptr = (u32 *)(SUNXI_DRAM_PHY0_BASE + 0x654);
> for (i = 0; i < 9; i++) {
> - writel_relaxed(0x14, ptr);
> - writel_relaxed(0x14, ptr + 0x30);
> + writel_relaxed(val, ptr);
> + writel_relaxed(val, ptr + 0x30);
> ptr += 2;
> }
> - writel_relaxed(0x1c, SUNXI_DRAM_PHY0_BASE + 0x6ac);
> - writel_relaxed(0x1c, SUNXI_DRAM_PHY0_BASE + 0x76c);
> - writel_relaxed(0x1c, SUNXI_DRAM_PHY0_BASE + 0x69c);
> - writel_relaxed(0x1c, SUNXI_DRAM_PHY0_BASE + 0x75c);
> +
> + if (para->tpr10 & TPR10_UNKNOWN_FEAT3)
> + val = (para->odt_en >> 11) & 0x1e;
> + else
> + val = (para->tpr12 >> 27) & 0x1e;
> +
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x6ac);
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x76c);
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x69c);
> + writel_relaxed(val, SUNXI_DRAM_PHY0_BASE + 0x75c);
>
> dmb();
>
> @@ -1021,7 +1109,10 @@ unsigned long sunxi_dram_init(void)
> .dx_odt = CONFIG_DRAM_SUN50I_H616_DX_ODT,
> .dx_dri = CONFIG_DRAM_SUN50I_H616_DX_DRI,
> .ca_dri = CONFIG_DRAM_SUN50I_H616_CA_DRI,
> + .odt_en = CONFIG_DRAM_SUN50I_H616_ODT_EN,
> .tpr10 = CONFIG_DRAM_SUN50I_H616_TPR10,
> + .tpr11 = CONFIG_DRAM_SUN50I_H616_TPR11,
> + .tpr12 = CONFIG_DRAM_SUN50I_H616_TPR12,
> };
> unsigned long size;
>
next prev parent reply other threads:[~2023-01-04 0:41 UTC|newest]
Thread overview: 29+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-12-11 16:32 [PATCH 0/8] sunxi: Update H616 DRAM driver Jernej Skrabec
2022-12-11 16:32 ` [PATCH 1/8] sunxi: Fix write to H616 DRAM CR register Jernej Skrabec
2023-01-04 0:35 ` Andre Przywara
2022-12-11 16:32 ` [PATCH 2/8] sunxi: cosmetic: Fix H616 DRAM driver code style Jernej Skrabec
2023-01-04 0:36 ` Andre Przywara
2022-12-11 16:32 ` [PATCH 3/8] sunxi: parameterize H616 DRAM ODT values Jernej Skrabec
2023-01-04 0:36 ` Andre Przywara
2022-12-11 16:32 ` [PATCH 4/8] sunxi: Convert H616 DRAM options to single setting Jernej Skrabec
2022-12-12 17:50 ` Andre Przywara
2022-12-13 16:23 ` Jernej Škrabec
2022-12-13 16:51 ` Andre Przywara
2022-12-13 17:08 ` Jernej Škrabec
2022-12-11 16:32 ` [PATCH 5/8] sunxi: Always configure ODT on H616 DRAM Jernej Skrabec
2023-01-04 0:37 ` Andre Przywara
2023-01-04 21:12 ` Jernej Škrabec
2022-12-11 16:32 ` [PATCH 6/8] sunxi: Make bit delay function in H616 DRAM code void Jernej Skrabec
2023-01-04 0:37 ` Andre Przywara
2022-12-11 16:32 ` [PATCH 7/8] sunxi: Parameterize bit delay code in H616 DRAM driver Jernej Skrabec
2023-01-04 0:37 ` Andre Przywara [this message]
2023-01-04 21:28 ` Jernej Škrabec
2022-12-11 16:32 ` [PATCH 8/8] sunxi: Parameterize H616 DRAM code some more Jernej Skrabec
2022-12-11 18:33 ` Jernej Škrabec
2023-01-04 0:38 ` Andre Przywara
2023-01-04 21:30 ` Jernej Škrabec
2022-12-12 1:04 ` [PATCH 0/8] sunxi: Update H616 DRAM driver Andre Przywara
2022-12-12 16:14 ` Jernej Škrabec
2023-01-04 0:47 ` Andre Przywara
2023-01-04 21:02 ` Jernej Škrabec
2023-01-04 23:21 ` Andre Przywara
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20230104003747.2d5a16f6@slackpad.lan \
--to=andre.przywara@arm.com \
--cc=jagan@amarulasolutions.com \
--cc=jernej.skrabec@gmail.com \
--cc=u-boot@lists.denx.de \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox