[PATCH 0/3] Add dw_mmc support for rk3576

[PATCH 0/3] Add dw_mmc support for rk3576

[Detlev Casanova]

The SD card controller on the rk3576 SoC uses a new tuning version that is
capable of using pre-boot tuning information.

Also, it stores the phase settings into the dw_mmc controller, so the code
has to be adapted to implement that.

Detlev Casanova (1):
  dt-bindings: pinctrl: Add rk3576 dw-mshc bindings

Shawn Lin (2):
  mmc: dw_mmc-rockchip: Add v2 tuning support
  mmc: dw_mmc-rockchip: Add internal phase support

 .../bindings/mmc/rockchip-dw-mshc.yaml        |  11 +
 drivers/mmc/host/dw_mmc-rockchip.c            | 224 +++++++++++++++++-
 2 files changed, 226 insertions(+), 9 deletions(-)

-- 
2.46.0

[PATCH 1/3] dt-bindings: pinctrl: Add rk3576 dw-mshc bindings

[Detlev Casanova]

Add the compatible string for rockchip,rk3576-dw-mshc and add support
for the rockchip,use-v2-tuning flag, a new feature of this core.

Signed-off-by: Detlev Casanova <detlev.casanova@collabora.com>
---
 .../devicetree/bindings/mmc/rockchip-dw-mshc.yaml     | 11 +++++++++++
 1 file changed, 11 insertions(+)

diff --git a/Documentation/devicetree/bindings/mmc/rockchip-dw-mshc.yaml b/Documentation/devicetree/bindings/mmc/rockchip-dw-mshc.yaml
index 211cd0b0bc5f3..dd8d1e773bb38 100644
--- a/Documentation/devicetree/bindings/mmc/rockchip-dw-mshc.yaml
+++ b/Documentation/devicetree/bindings/mmc/rockchip-dw-mshc.yaml
@@ -39,6 +39,7 @@ properties:
               - rockchip,rk3368-dw-mshc
               - rockchip,rk3399-dw-mshc
               - rockchip,rk3568-dw-mshc
+              - rockchip,rk3576-dw-mshc
               - rockchip,rk3588-dw-mshc
               - rockchip,rv1108-dw-mshc
               - rockchip,rv1126-dw-mshc
@@ -95,6 +96,16 @@ properties:
       If not specified, the host will do tuning for 360 times,
       namely tuning for each degree.
 
+  rockchip,use-v2-tuning:
+    $ref: /schemas/types.yaml#/definitions/flag
+    description:
+      If present, use tuning version 2.
+      v2 tuning will inherit pre-stage loader's phase settings for the first
+      time, and do re-tune if necessary.
+      Re-tune will still try the rough degrees, for instance, 90, 180, 270,
+      360 but continue to do the fine tuning if sample window isn't good
+      enough.
+
 required:
   - compatible
   - reg
-- 
2.46.0

[PATCH 2/3] mmc: dw_mmc-rockchip: Add v2 tuning support

[Detlev Casanova]

From: Shawn Lin <shawn.lin@rock-chips.com>

v2 tuning will inherit pre-stage loader's phase settings for the first
time, and do re-tune if necessary.
Re-tune will still try the rough degrees, for instance, 90, 180, 270,
360 but continue to do the fine tuning if sample window isn't good
enough.

Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Signed-off-by: Detlev Casanova <detlev.casanova@collabora.com>
---
 drivers/mmc/host/dw_mmc-rockchip.c | 48 ++++++++++++++++++++++++++++++
 1 file changed, 48 insertions(+)

diff --git a/drivers/mmc/host/dw_mmc-rockchip.c b/drivers/mmc/host/dw_mmc-rockchip.c
index b07190ba4b7ac..c47b93d99a1b3 100644
--- a/drivers/mmc/host/dw_mmc-rockchip.c
+++ b/drivers/mmc/host/dw_mmc-rockchip.c
@@ -24,6 +24,8 @@ struct dw_mci_rockchip_priv_data {
 	struct clk		*sample_clk;
 	int			default_sample_phase;
 	int			num_phases;
+	bool			use_v2_tuning;
+	int			last_degree;
 };
 
 static void dw_mci_rk3288_set_ios(struct dw_mci *host, struct mmc_ios *ios)
@@ -134,6 +136,42 @@ static void dw_mci_rk3288_set_ios(struct dw_mci *host, struct mmc_ios *ios)
 #define TUNING_ITERATION_TO_PHASE(i, num_phases) \
 		(DIV_ROUND_UP((i) * 360, num_phases))
 
+static int dw_mci_v2_execute_tuning(struct dw_mci_slot *slot, u32 opcode)
+{
+	struct dw_mci *host = slot->host;
+	struct dw_mci_rockchip_priv_data *priv = host->priv;
+	struct mmc_host *mmc = slot->mmc;
+	u32 degrees[4] = {90, 180, 270, 360};
+	int i;
+	static bool inherit = true;
+
+	if (inherit) {
+		inherit = false;
+		i = clk_get_phase(priv->sample_clk) / 90 - 1;
+		goto done;
+	}
+
+	/* v2 only support 4 degrees in theory */
+	for (i = 0; i < ARRAY_SIZE(degrees); i++) {
+		if (degrees[i] == priv->last_degree)
+			continue;
+
+		clk_set_phase(priv->sample_clk, degrees[i]);
+		if (!mmc_send_tuning(mmc, opcode, NULL))
+			break;
+	}
+
+	if (i == ARRAY_SIZE(degrees)) {
+		dev_warn(host->dev, "All phases bad!");
+		return -EIO;
+	}
+
+done:
+	dev_info(host->dev, "Successfully tuned phase to %d\n", degrees[i]);
+	priv->last_degree = degrees[i];
+	return 0;
+}
+
 static int dw_mci_rk3288_execute_tuning(struct dw_mci_slot *slot, u32 opcode)
 {
 	struct dw_mci *host = slot->host;
@@ -157,6 +195,13 @@ static int dw_mci_rk3288_execute_tuning(struct dw_mci_slot *slot, u32 opcode)
 		return -EIO;
 	}
 
+	if (priv->use_v2_tuning) {
+		ret = dw_mci_v2_execute_tuning(slot, opcode);
+		if (!ret)
+			return 0;
+		/* Otherwise we continue using fine tuning */
+	}
+
 	ranges = kmalloc_array(priv->num_phases / 2 + 1,
 			       sizeof(*ranges), GFP_KERNEL);
 	if (!ranges)
@@ -277,6 +322,9 @@ static int dw_mci_rk3288_parse_dt(struct dw_mci *host)
 					&priv->default_sample_phase))
 		priv->default_sample_phase = 0;
 
+	if (of_property_read_bool(np, "rockchip,use-v2-tuning"))
+		priv->use_v2_tuning = true;
+
 	priv->drv_clk = devm_clk_get(host->dev, "ciu-drive");
 	if (IS_ERR(priv->drv_clk))
 		dev_dbg(host->dev, "ciu-drive not available\n");
-- 
2.46.0

[PATCH 3/3] mmc: dw_mmc-rockchip: Add internal phase support

[Detlev Casanova]

From: Shawn Lin <shawn.lin@rock-chips.com>

Some Rockchip devices put the phase settings into the dw_mmc controller.

The feature is implemented in devices where the USRID register contains
0x20230002.

Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Signed-off-by: Detlev Casanova <detlev.casanova@collabora.com>
---
 drivers/mmc/host/dw_mmc-rockchip.c | 180 +++++++++++++++++++++++++++--
 1 file changed, 169 insertions(+), 11 deletions(-)

diff --git a/drivers/mmc/host/dw_mmc-rockchip.c b/drivers/mmc/host/dw_mmc-rockchip.c
index c47b93d99a1b3..2f474e7eb21dc 100644
--- a/drivers/mmc/host/dw_mmc-rockchip.c
+++ b/drivers/mmc/host/dw_mmc-rockchip.c
@@ -16,6 +16,17 @@
 #include "dw_mmc-pltfm.h"
 
 #define RK3288_CLKGEN_DIV	2
+#define USRID_INTER_PHASE	0x20230001
+#define SDMMC_TIMING_CON0	0x130
+#define SDMMC_TIMING_CON1	0x134
+#define ROCKCHIP_MMC_DELAY_SEL BIT(10)
+#define ROCKCHIP_MMC_DEGREE_MASK 0x3
+#define ROCKCHIP_MMC_DELAYNUM_OFFSET 2
+#define ROCKCHIP_MMC_DELAYNUM_MASK (0xff << ROCKCHIP_MMC_DELAYNUM_OFFSET)
+#define PSECS_PER_SEC 1000000000000LL
+#define ROCKCHIP_MMC_DELAY_ELEMENT_PSEC 60
+#define HIWORD_UPDATE(val, mask, shift) \
+		((val) << (shift) | (mask) << ((shift) + 16))
 
 static const unsigned int freqs[] = { 100000, 200000, 300000, 400000 };
 
@@ -25,9 +36,121 @@ struct dw_mci_rockchip_priv_data {
 	int			default_sample_phase;
 	int			num_phases;
 	bool			use_v2_tuning;
+	int			usrid;
 	int			last_degree;
 };
 
+/*
+ * Each fine delay is between 44ps-77ps. Assume each fine delay is 60ps to
+ * simplify calculations. So 45degs could be anywhere between 33deg and 57.8deg.
+ */
+static int rockchip_mmc_get_phase(struct dw_mci *host, bool sample)
+{
+	unsigned long rate = clk_get_rate(host->ciu_clk);
+	u32 raw_value;
+	u16 degrees;
+	u32 delay_num = 0;
+
+	/* Constant signal, no measurable phase shift */
+	if (!rate)
+		return 0;
+
+	if (sample)
+		raw_value = mci_readl(host, TIMING_CON1) >> 1;
+	else
+		raw_value = mci_readl(host, TIMING_CON0) >> 1;
+
+	degrees = (raw_value & ROCKCHIP_MMC_DEGREE_MASK) * 90;
+
+	if (raw_value & ROCKCHIP_MMC_DELAY_SEL) {
+		/* degrees/delaynum * 1000000 */
+		unsigned long factor = (ROCKCHIP_MMC_DELAY_ELEMENT_PSEC / 10) *
+					36 * (rate / 10000);
+
+		delay_num = (raw_value & ROCKCHIP_MMC_DELAYNUM_MASK);
+		delay_num >>= ROCKCHIP_MMC_DELAYNUM_OFFSET;
+		degrees += DIV_ROUND_CLOSEST(delay_num * factor, 1000000);
+	}
+
+	return degrees % 360;
+}
+
+static int rockchip_mmc_set_phase(struct dw_mci *host, bool sample, int degrees)
+{
+	unsigned long rate = clk_get_rate(host->ciu_clk);
+	u8 nineties, remainder;
+	u8 delay_num;
+	u32 raw_value;
+	u32 delay;
+
+	/*
+	 * The below calculation is based on the output clock from
+	 * MMC host to the card, which expects the phase clock inherits
+	 * the clock rate from its parent, namely the output clock
+	 * provider of MMC host. However, things may go wrong if
+	 * (1) It is orphan.
+	 * (2) It is assigned to the wrong parent.
+	 *
+	 * This check help debug the case (1), which seems to be the
+	 * most likely problem we often face and which makes it difficult
+	 * for people to debug unstable mmc tuning results.
+	 */
+	if (!rate) {
+		dev_err(host->dev, "%s: invalid clk rate\n", __func__);
+		return -EINVAL;
+	}
+
+	nineties = degrees / 90;
+	remainder = (degrees % 90);
+
+	/*
+	 * Due to the inexact nature of the "fine" delay, we might
+	 * actually go non-monotonic.  We don't go _too_ monotonic
+	 * though, so we should be OK.  Here are options of how we may
+	 * work:
+	 *
+	 * Ideally we end up with:
+	 *   1.0, 2.0, ..., 69.0, 70.0, ...,  89.0, 90.0
+	 *
+	 * On one extreme (if delay is actually 44ps):
+	 *   .73, 1.5, ..., 50.6, 51.3, ...,  65.3, 90.0
+	 * The other (if delay is actually 77ps):
+	 *   1.3, 2.6, ..., 88.6. 89.8, ..., 114.0, 90
+	 *
+	 * It's possible we might make a delay that is up to 25
+	 * degrees off from what we think we're making.  That's OK
+	 * though because we should be REALLY far from any bad range.
+	 */
+
+	/*
+	 * Convert to delay; do a little extra work to make sure we
+	 * don't overflow 32-bit / 64-bit numbers.
+	 */
+	delay = 10000000; /* PSECS_PER_SEC / 10000 / 10 */
+	delay *= remainder;
+	delay = DIV_ROUND_CLOSEST(delay,
+			(rate / 1000) * 36 *
+				(ROCKCHIP_MMC_DELAY_ELEMENT_PSEC / 10));
+
+	delay_num = (u8) min_t(u32, delay, 255);
+
+	raw_value = delay_num ? ROCKCHIP_MMC_DELAY_SEL : 0;
+	raw_value |= delay_num << ROCKCHIP_MMC_DELAYNUM_OFFSET;
+	raw_value |= nineties;
+
+	if (sample)
+		mci_writel(host, TIMING_CON1, HIWORD_UPDATE(raw_value, 0x07ff, 1));
+	else
+		mci_writel(host, TIMING_CON0, HIWORD_UPDATE(raw_value, 0x07ff, 1));
+
+	dev_dbg(host->dev, "set %s_phase(%d) delay_nums=%u actual_degrees=%d\n",
+		sample ? "sample" : "drv", degrees, delay_num,
+		rockchip_mmc_get_phase(host, sample)
+	);
+
+	return 0;
+}
+
 static void dw_mci_rk3288_set_ios(struct dw_mci *host, struct mmc_ios *ios)
 {
 	struct dw_mci_rockchip_priv_data *priv = host->priv;
@@ -65,8 +188,12 @@ static void dw_mci_rk3288_set_ios(struct dw_mci *host, struct mmc_ios *ios)
 	}
 
 	/* Make sure we use phases which we can enumerate with */
-	if (!IS_ERR(priv->sample_clk) && ios->timing <= MMC_TIMING_SD_HS)
-		clk_set_phase(priv->sample_clk, priv->default_sample_phase);
+	if (!IS_ERR(priv->sample_clk) && ios->timing <= MMC_TIMING_SD_HS) {
+		if (priv->usrid == USRID_INTER_PHASE)
+			rockchip_mmc_set_phase(host, true, priv->default_sample_phase);
+		else
+			clk_set_phase(priv->sample_clk, priv->default_sample_phase);
+	}
 
 	/*
 	 * Set the drive phase offset based on speed mode to achieve hold times.
@@ -129,7 +256,10 @@ static void dw_mci_rk3288_set_ios(struct dw_mci *host, struct mmc_ios *ios)
 			break;
 		}
 
-		clk_set_phase(priv->drv_clk, phase);
+		if (priv->usrid == USRID_INTER_PHASE)
+			rockchip_mmc_set_phase(host, false, phase);
+		else
+			clk_set_phase(priv->drv_clk, phase);
 	}
 }
 
@@ -147,7 +277,10 @@ static int dw_mci_v2_execute_tuning(struct dw_mci_slot *slot, u32 opcode)
 
 	if (inherit) {
 		inherit = false;
-		i = clk_get_phase(priv->sample_clk) / 90 - 1;
+		if (priv->usrid == USRID_INTER_PHASE)
+			i = rockchip_mmc_get_phase(host, true) / 90;
+		else
+			i = clk_get_phase(priv->sample_clk) / 90 - 1;
 		goto done;
 	}
 
@@ -156,7 +289,12 @@ static int dw_mci_v2_execute_tuning(struct dw_mci_slot *slot, u32 opcode)
 		if (degrees[i] == priv->last_degree)
 			continue;
 
-		clk_set_phase(priv->sample_clk, degrees[i]);
+		u32 degree = degrees[i] + priv->last_degree + 90;
+		degree = degree % 360;
+		if (priv->usrid == USRID_INTER_PHASE)
+			rockchip_mmc_set_phase(host, true, degree);
+		else
+			clk_set_phase(priv->sample_clk, degree);
 		if (!mmc_send_tuning(mmc, opcode, NULL))
 			break;
 	}
@@ -209,8 +347,15 @@ static int dw_mci_rk3288_execute_tuning(struct dw_mci_slot *slot, u32 opcode)
 
 	/* Try each phase and extract good ranges */
 	for (i = 0; i < priv->num_phases; ) {
-		clk_set_phase(priv->sample_clk,
-			      TUNING_ITERATION_TO_PHASE(i, priv->num_phases));
+		/* Cannot guarantee any phases larger than 270 would work well */
+		if (TUNING_ITERATION_TO_PHASE(i, priv->num_phases) > 270)
+			break;
+		if (priv->usrid == USRID_INTER_PHASE)
+			rockchip_mmc_set_phase(host, true,
+				TUNING_ITERATION_TO_PHASE(i, priv->num_phases));
+		else
+			clk_set_phase(priv->sample_clk,
+				TUNING_ITERATION_TO_PHASE(i, priv->num_phases));
 
 		v = !mmc_send_tuning(mmc, opcode, NULL);
 
@@ -256,7 +401,10 @@ static int dw_mci_rk3288_execute_tuning(struct dw_mci_slot *slot, u32 opcode)
 	}
 
 	if (ranges[0].start == 0 && ranges[0].end == priv->num_phases - 1) {
-		clk_set_phase(priv->sample_clk, priv->default_sample_phase);
+		if (priv->usrid == USRID_INTER_PHASE)
+			rockchip_mmc_set_phase(host, true, priv->default_sample_phase);
+		else
+			clk_set_phase(priv->sample_clk, priv->default_sample_phase);
 		dev_info(host->dev, "All phases work, using default phase %d.",
 			 priv->default_sample_phase);
 		goto free;
@@ -296,9 +444,12 @@ static int dw_mci_rk3288_execute_tuning(struct dw_mci_slot *slot, u32 opcode)
 	dev_info(host->dev, "Successfully tuned phase to %d\n",
 		 TUNING_ITERATION_TO_PHASE(middle_phase, priv->num_phases));
 
-	clk_set_phase(priv->sample_clk,
-		      TUNING_ITERATION_TO_PHASE(middle_phase,
-						priv->num_phases));
+	if (priv->usrid == USRID_INTER_PHASE)
+		rockchip_mmc_set_phase(host, true, TUNING_ITERATION_TO_PHASE(middle_phase,
+                                                priv->num_phases));
+	else
+		clk_set_phase(priv->sample_clk, TUNING_ITERATION_TO_PHASE(middle_phase,
+                                                priv->num_phases));
 
 free:
 	kfree(ranges);
@@ -341,6 +492,7 @@ static int dw_mci_rk3288_parse_dt(struct dw_mci *host)
 static int dw_mci_rockchip_init(struct dw_mci *host)
 {
 	int ret, i;
+	struct dw_mci_rockchip_priv_data *priv = host->priv;
 
 	/* It is slot 8 on Rockchip SoCs */
 	host->sdio_id0 = 8;
@@ -364,6 +516,12 @@ static int dw_mci_rockchip_init(struct dw_mci *host)
 			dev_warn(host->dev, "no valid minimum freq: %d\n", ret);
 	}
 
+	priv->usrid = mci_readl(host, USRID);
+	if (priv->usrid == USRID_INTER_PHASE) {
+		priv->sample_clk = NULL;
+		priv->drv_clk = NULL;
+	}
+
 	return 0;
 }
 
-- 
2.46.0

Re: [PATCH 1/3] dt-bindings: pinctrl: Add rk3576 dw-mshc bindings

[Detlev Casanova]

> [PATCH 1/3] dt-bindings: pinctrl: Add rk3576 dw-mshc bindings

s/pinctrl/mmc

On Friday, 2 August 2024 11:31:27 EDT Detlev Casanova wrote:
> Add the compatible string for rockchip,rk3576-dw-mshc and add support
> for the rockchip,use-v2-tuning flag, a new feature of this core.
> 
> Signed-off-by: Detlev Casanova <detlev.casanova@collabora.com>
> ---
>  .../devicetree/bindings/mmc/rockchip-dw-mshc.yaml     | 11 +++++++++++
>  1 file changed, 11 insertions(+)
> 
> diff --git a/Documentation/devicetree/bindings/mmc/rockchip-dw-mshc.yaml
> b/Documentation/devicetree/bindings/mmc/rockchip-dw-mshc.yaml index
> 211cd0b0bc5f3..dd8d1e773bb38 100644
> --- a/Documentation/devicetree/bindings/mmc/rockchip-dw-mshc.yaml
> +++ b/Documentation/devicetree/bindings/mmc/rockchip-dw-mshc.yaml
> @@ -39,6 +39,7 @@ properties:
>                - rockchip,rk3368-dw-mshc
>                - rockchip,rk3399-dw-mshc
>                - rockchip,rk3568-dw-mshc
> +              - rockchip,rk3576-dw-mshc
>                - rockchip,rk3588-dw-mshc
>                - rockchip,rv1108-dw-mshc
>                - rockchip,rv1126-dw-mshc
> @@ -95,6 +96,16 @@ properties:
>        If not specified, the host will do tuning for 360 times,
>        namely tuning for each degree.
> 
> +  rockchip,use-v2-tuning:
> +    $ref: /schemas/types.yaml#/definitions/flag
> +    description:
> +      If present, use tuning version 2.
> +      v2 tuning will inherit pre-stage loader's phase settings for the
> first +      time, and do re-tune if necessary.
> +      Re-tune will still try the rough degrees, for instance, 90, 180, 270,
> +      360 but continue to do the fine tuning if sample window isn't good +
>      enough.
> +
>  required:
>    - compatible
>    - reg

Re: [PATCH 1/3] dt-bindings: pinctrl: Add rk3576 dw-mshc bindings

[Krzysztof Kozlowski]

On 02/08/2024 17:31, Detlev Casanova wrote:
> Add the compatible string for rockchip,rk3576-dw-mshc and add support
> for the rockchip,use-v2-tuning flag, a new feature of this core.
> 
> Signed-off-by: Detlev Casanova <detlev.casanova@collabora.com>

A nit, subject: drop second/last, redundant "bindings". The
"dt-bindings" prefix is already stating that these are bindings.
See also:
https://elixir.bootlin.com/linux/v6.7-rc8/source/Documentation/devicetree/bindings/submitting-patches.rst#L18

> ---
>  .../devicetree/bindings/mmc/rockchip-dw-mshc.yaml     | 11 +++++++++++
>  1 file changed, 11 insertions(+)
> 
> diff --git a/Documentation/devicetree/bindings/mmc/rockchip-dw-mshc.yaml b/Documentation/devicetree/bindings/mmc/rockchip-dw-mshc.yaml
> index 211cd0b0bc5f3..dd8d1e773bb38 100644
> --- a/Documentation/devicetree/bindings/mmc/rockchip-dw-mshc.yaml
> +++ b/Documentation/devicetree/bindings/mmc/rockchip-dw-mshc.yaml
> @@ -39,6 +39,7 @@ properties:
>                - rockchip,rk3368-dw-mshc
>                - rockchip,rk3399-dw-mshc
>                - rockchip,rk3568-dw-mshc
> +              - rockchip,rk3576-dw-mshc
>                - rockchip,rk3588-dw-mshc
>                - rockchip,rv1108-dw-mshc
>                - rockchip,rv1126-dw-mshc
> @@ -95,6 +96,16 @@ properties:
>        If not specified, the host will do tuning for 360 times,
>        namely tuning for each degree.
>  
> +  rockchip,use-v2-tuning:
> +    $ref: /schemas/types.yaml#/definitions/flag
> +    description:
> +      If present, use tuning version 2.
> +      v2 tuning will inherit pre-stage loader's phase settings for the first
> +      time, and do re-tune if necessary.
> +      Re-tune will still try the rough degrees, for instance, 90, 180, 270,
> +      360 but continue to do the fine tuning if sample window isn't good
> +      enough.

You described the desired Linux feature or behavior, not the actual
hardware. The bindings are about the latter, so instead you need to
rephrase the property and its description to match actual hardware
capabilities/features/configuration etc.

Best regards,
Krzysztof