mmc: mmci_sdmmc: Add execute tuning with delay block

The hardware delay block is used to align the sampling clock on the data
received by SDMMC. It is mandatory for SDMMC to support the SDR104 mode.
The delay block is used to generate an output clock which is dephased from
the input clock.  The phase of the output clock must be programmed by the
execute tuning interface.
Signed-off-by: Ludovic Barre <ludovic.barre@st.com>
Link: https://lore.kernel.org/r/20200128090636.13689-7-ludovic.barre@st.comSigned-off-by: Ulf Hansson <ulf.hansson@linaro.org>

drivers/mmc/host/mmci_stm32_sdmmc.c

@@ -3,10 +3,13 @@
  * Copyright (C) STMicroelectronics 2018 - All Rights Reserved
  * Author: Ludovic.barre@st.com for STMicroelectronics.
  */
+#include <linux/bitfield.h>
 #include <linux/delay.h>
 #include <linux/dma-mapping.h>
+#include <linux/iopoll.h>
 #include <linux/mmc/host.h>
 #include <linux/mmc/card.h>
+#include <linux/of_address.h>
 #include <linux/reset.h>
 #include <linux/scatterlist.h>
 #include "mmci.h"
@@ -14,6 +17,22 @@
 #define SDMMC_LLI_BUF_LEN	PAGE_SIZE
 #define SDMMC_IDMA_BURST	BIT(MMCI_STM32_IDMABNDT_SHIFT)
 
+#define DLYB_CR			0x0
+#define DLYB_CR_DEN		BIT(0)
+#define DLYB_CR_SEN		BIT(1)
+
+#define DLYB_CFGR		0x4
+#define DLYB_CFGR_SEL_MASK	GENMASK(3, 0)
+#define DLYB_CFGR_UNIT_MASK	GENMASK(14, 8)
+#define DLYB_CFGR_LNG_MASK	GENMASK(27, 16)
+#define DLYB_CFGR_LNGF		BIT(31)
+
+#define DLYB_NB_DELAY		11
+#define DLYB_CFGR_SEL_MAX	(DLYB_NB_DELAY + 1)
+#define DLYB_CFGR_UNIT_MAX	127
+
+#define DLYB_LNG_TIMEOUT_US	1000
+
 struct sdmmc_lli_desc {
 	u32 idmalar;
 	u32 idmabase;
@@ -25,6 +44,12 @@ struct sdmmc_idma {
 	void *sg_cpu;
 };
 
+struct sdmmc_dlyb {
+	void __iomem *base;
+	u32 unit;
+	u32 max;
+};
+
 static int sdmmc_idma_validate_data(struct mmci_host *host,
 				    struct mmc_data *data)
 {
@@ -226,12 +251,24 @@ static void mmci_sdmmc_set_clkreg(struct mmci_host *host, unsigned int desired)
 	mmci_write_clkreg(host, clk);
 }
 
+static void sdmmc_dlyb_input_ck(struct sdmmc_dlyb *dlyb)
+{
+	if (!dlyb || !dlyb->base)
+		return;
+
+	/* Output clock = Input clock */
+	writel_relaxed(0, dlyb->base + DLYB_CR);
+}
+
 static void mmci_sdmmc_set_pwrreg(struct mmci_host *host, unsigned int pwr)
 {
 	struct mmc_ios ios = host->mmc->ios;
+	struct sdmmc_dlyb *dlyb = host->variant_priv;
 
 	pwr = host->pwr_reg_add;
 
+	sdmmc_dlyb_input_ck(dlyb);
+
 	if (ios.power_mode == MMC_POWER_OFF) {
 		/* Only a reset could power-off sdmmc */
 		reset_control_assert(host->rst);
@@ -323,6 +360,102 @@ static bool sdmmc_busy_complete(struct mmci_host *host, u32 status, u32 err_msk)
 	return true;
 }
 
+static void sdmmc_dlyb_set_cfgr(struct sdmmc_dlyb *dlyb,
+				int unit, int phase, bool sampler)
+{
+	u32 cfgr;
+
+	writel_relaxed(DLYB_CR_SEN | DLYB_CR_DEN, dlyb->base + DLYB_CR);
+
+	cfgr = FIELD_PREP(DLYB_CFGR_UNIT_MASK, unit) |
+	       FIELD_PREP(DLYB_CFGR_SEL_MASK, phase);
+	writel_relaxed(cfgr, dlyb->base + DLYB_CFGR);
+
+	if (!sampler)
+		writel_relaxed(DLYB_CR_DEN, dlyb->base + DLYB_CR);
+}
+
+static int sdmmc_dlyb_lng_tuning(struct mmci_host *host)
+{
+	struct sdmmc_dlyb *dlyb = host->variant_priv;
+	u32 cfgr;
+	int i, lng, ret;
+
+	for (i = 0; i <= DLYB_CFGR_UNIT_MAX; i++) {
+		sdmmc_dlyb_set_cfgr(dlyb, i, DLYB_CFGR_SEL_MAX, true);
+
+		ret = readl_relaxed_poll_timeout(dlyb->base + DLYB_CFGR, cfgr,
+						 (cfgr & DLYB_CFGR_LNGF),
+						 1, DLYB_LNG_TIMEOUT_US);
+		if (ret) {
+			dev_warn(mmc_dev(host->mmc),
+				 "delay line cfg timeout unit:%d cfgr:%d\n",
+				 i, cfgr);
+			continue;
+		}
+
+		lng = FIELD_GET(DLYB_CFGR_LNG_MASK, cfgr);
+		if (lng < BIT(DLYB_NB_DELAY) && lng > 0)
+			break;
+	}
+
+	if (i > DLYB_CFGR_UNIT_MAX)
+		return -EINVAL;
+
+	dlyb->unit = i;
+	dlyb->max = __fls(lng);
+
+	return 0;
+}
+
+static int sdmmc_dlyb_phase_tuning(struct mmci_host *host, u32 opcode)
+{
+	struct sdmmc_dlyb *dlyb = host->variant_priv;
+	int cur_len = 0, max_len = 0, end_of_len = 0;
+	int phase;
+
+	for (phase = 0; phase <= dlyb->max; phase++) {
+		sdmmc_dlyb_set_cfgr(dlyb, dlyb->unit, phase, false);
+
+		if (mmc_send_tuning(host->mmc, opcode, NULL)) {
+			cur_len = 0;
+		} else {
+			cur_len++;
+			if (cur_len > max_len) {
+				max_len = cur_len;
+				end_of_len = phase;
+			}
+		}
+	}
+
+	if (!max_len) {
+		dev_err(mmc_dev(host->mmc), "no tuning point found\n");
+		return -EINVAL;
+	}
+
+	phase = end_of_len - max_len / 2;
+	sdmmc_dlyb_set_cfgr(dlyb, dlyb->unit, phase, false);
+
+	dev_dbg(mmc_dev(host->mmc), "unit:%d max_dly:%d phase:%d\n",
+		dlyb->unit, dlyb->max, phase);
+
+	return 0;
+}
+
+static int sdmmc_execute_tuning(struct mmc_host *mmc, u32 opcode)
+{
+	struct mmci_host *host = mmc_priv(mmc);
+	struct sdmmc_dlyb *dlyb = host->variant_priv;
+
+	if (!dlyb || !dlyb->base)
+		return -EINVAL;
+
+	if (sdmmc_dlyb_lng_tuning(host))
+		return -EINVAL;
+
+	return sdmmc_dlyb_phase_tuning(host, opcode);
+}
+
 static struct mmci_host_ops sdmmc_variant_ops = {
 	.validate_data = sdmmc_idma_validate_data,
 	.prep_data = sdmmc_idma_prep_data,
@@ -338,5 +471,21 @@ static struct mmci_host_ops sdmmc_variant_ops = {
 
 void sdmmc_variant_init(struct mmci_host *host)
 {
+	struct device_node *np = host->mmc->parent->of_node;
+	void __iomem *base_dlyb;
+	struct sdmmc_dlyb *dlyb;
+
 	host->ops = &sdmmc_variant_ops;
+
+	base_dlyb = devm_of_iomap(mmc_dev(host->mmc), np, 1, NULL);
+	if (IS_ERR(base_dlyb))
+		return;
+
+	dlyb = devm_kzalloc(mmc_dev(host->mmc), sizeof(*dlyb), GFP_KERNEL);
+	if (!dlyb)
+		return;
+
+	dlyb->base = base_dlyb;
+	host->variant_priv = dlyb;
+	host->mmc_ops->execute_tuning = sdmmc_execute_tuning;
 }