21 files changed, 591 insertions, 374 deletions

diff --git a/drivers/mmc/host/au1xmmc.c b/drivers/mmc/host/au1xmmc.c
index 0acc237843f7..a9a0837153d8 100644
--- a/drivers/mmc/host/au1xmmc.c
+++ b/drivers/mmc/host/au1xmmc.c
@@ -969,8 +969,10 @@ static int au1xmmc_probe(struct platform_device *pdev)
 	}
 
 	host->irq = platform_get_irq(pdev, 0);
-	if (host->irq < 0)
+	if (host->irq < 0) {
+		ret = host->irq;
 		goto out3;
+	}
 
 	mmc->ops = &au1xmmc_ops;
 
diff --git a/drivers/mmc/host/dw_mmc-exynos.c b/drivers/mmc/host/dw_mmc-exynos.c
index c2dd29ef45c6..ca5be4445ae0 100644
--- a/drivers/mmc/host/dw_mmc-exynos.c
+++ b/drivers/mmc/host/dw_mmc-exynos.c
@@ -28,6 +28,7 @@ enum dw_mci_exynos_type {
 	DW_MCI_TYPE_EXYNOS5420_SMU,
 	DW_MCI_TYPE_EXYNOS7,
 	DW_MCI_TYPE_EXYNOS7_SMU,
+	DW_MCI_TYPE_ARTPEC8,
 };
 
 /* Exynos implementation specific driver private data */
@@ -69,6 +70,9 @@ static struct dw_mci_exynos_compatible {
 	}, {
 		.compatible	= "samsung,exynos7-dw-mshc-smu",
 		.ctrl_type	= DW_MCI_TYPE_EXYNOS7_SMU,
+	}, {
+		.compatible	= "axis,artpec8-dw-mshc",
+		.ctrl_type	= DW_MCI_TYPE_ARTPEC8,
 	},
 };
 
@@ -81,7 +85,8 @@ static inline u8 dw_mci_exynos_get_ciu_div(struct dw_mci *host)
 	else if (priv->ctrl_type == DW_MCI_TYPE_EXYNOS4210)
 		return EXYNOS4210_FIXED_CIU_CLK_DIV;
 	else if (priv->ctrl_type == DW_MCI_TYPE_EXYNOS7 ||
-			priv->ctrl_type == DW_MCI_TYPE_EXYNOS7_SMU)
+			priv->ctrl_type == DW_MCI_TYPE_EXYNOS7_SMU ||
+			priv->ctrl_type == DW_MCI_TYPE_ARTPEC8)
 		return SDMMC_CLKSEL_GET_DIV(mci_readl(host, CLKSEL64)) + 1;
 	else
 		return SDMMC_CLKSEL_GET_DIV(mci_readl(host, CLKSEL)) + 1;
@@ -122,6 +127,11 @@ static int dw_mci_exynos_priv_init(struct dw_mci *host)
 				DQS_CTRL_GET_RD_DELAY(priv->saved_strobe_ctrl);
 	}
 
+	if (priv->ctrl_type == DW_MCI_TYPE_ARTPEC8) {
+		/* Quirk needed for the ARTPEC-8 SoC */
+		host->quirks |= DW_MMC_QUIRK_EXTENDED_TMOUT;
+	}
+
 	host->bus_hz /= (priv->ciu_div + 1);
 
 	return 0;
@@ -133,7 +143,8 @@ static void dw_mci_exynos_set_clksel_timing(struct dw_mci *host, u32 timing)
 	u32 clksel;
 
 	if (priv->ctrl_type == DW_MCI_TYPE_EXYNOS7 ||
-		priv->ctrl_type == DW_MCI_TYPE_EXYNOS7_SMU)
+		priv->ctrl_type == DW_MCI_TYPE_EXYNOS7_SMU ||
+		priv->ctrl_type == DW_MCI_TYPE_ARTPEC8)
 		clksel = mci_readl(host, CLKSEL64);
 	else
 		clksel = mci_readl(host, CLKSEL);
@@ -141,7 +152,8 @@ static void dw_mci_exynos_set_clksel_timing(struct dw_mci *host, u32 timing)
 	clksel = (clksel & ~SDMMC_CLKSEL_TIMING_MASK) | timing;
 
 	if (priv->ctrl_type == DW_MCI_TYPE_EXYNOS7 ||
-		priv->ctrl_type == DW_MCI_TYPE_EXYNOS7_SMU)
+		priv->ctrl_type == DW_MCI_TYPE_EXYNOS7_SMU ||
+		priv->ctrl_type == DW_MCI_TYPE_ARTPEC8)
 		mci_writel(host, CLKSEL64, clksel);
 	else
 		mci_writel(host, CLKSEL, clksel);
@@ -210,14 +222,16 @@ static int dw_mci_exynos_resume_noirq(struct device *dev)
 		return ret;
 
 	if (priv->ctrl_type == DW_MCI_TYPE_EXYNOS7 ||
-		priv->ctrl_type == DW_MCI_TYPE_EXYNOS7_SMU)
+		priv->ctrl_type == DW_MCI_TYPE_EXYNOS7_SMU ||
+		priv->ctrl_type == DW_MCI_TYPE_ARTPEC8)
 		clksel = mci_readl(host, CLKSEL64);
 	else
 		clksel = mci_readl(host, CLKSEL);
 
 	if (clksel & SDMMC_CLKSEL_WAKEUP_INT) {
 		if (priv->ctrl_type == DW_MCI_TYPE_EXYNOS7 ||
-			priv->ctrl_type == DW_MCI_TYPE_EXYNOS7_SMU)
+			priv->ctrl_type == DW_MCI_TYPE_EXYNOS7_SMU ||
+			priv->ctrl_type == DW_MCI_TYPE_ARTPEC8)
 			mci_writel(host, CLKSEL64, clksel);
 		else
 			mci_writel(host, CLKSEL, clksel);
@@ -238,7 +252,8 @@ static void dw_mci_exynos_config_hs400(struct dw_mci *host, u32 timing)
 	 * Not supported to configure register
 	 * related to HS400
 	 */
-	if (priv->ctrl_type < DW_MCI_TYPE_EXYNOS5420) {
+	if ((priv->ctrl_type < DW_MCI_TYPE_EXYNOS5420) ||
+		(priv->ctrl_type == DW_MCI_TYPE_ARTPEC8)) {
 		if (timing == MMC_TIMING_MMC_HS400)
 			dev_warn(host->dev,
 				 "cannot configure HS400, unsupported chipset\n");
@@ -394,7 +409,8 @@ static inline u8 dw_mci_exynos_get_clksmpl(struct dw_mci *host)
 	struct dw_mci_exynos_priv_data *priv = host->priv;
 
 	if (priv->ctrl_type == DW_MCI_TYPE_EXYNOS7 ||
-		priv->ctrl_type == DW_MCI_TYPE_EXYNOS7_SMU)
+		priv->ctrl_type == DW_MCI_TYPE_EXYNOS7_SMU ||
+		priv->ctrl_type == DW_MCI_TYPE_ARTPEC8)
 		return SDMMC_CLKSEL_CCLK_SAMPLE(mci_readl(host, CLKSEL64));
 	else
 		return SDMMC_CLKSEL_CCLK_SAMPLE(mci_readl(host, CLKSEL));
@@ -406,13 +422,15 @@ static inline void dw_mci_exynos_set_clksmpl(struct dw_mci *host, u8 sample)
 	struct dw_mci_exynos_priv_data *priv = host->priv;
 
 	if (priv->ctrl_type == DW_MCI_TYPE_EXYNOS7 ||
-		priv->ctrl_type == DW_MCI_TYPE_EXYNOS7_SMU)
+		priv->ctrl_type == DW_MCI_TYPE_EXYNOS7_SMU ||
+		priv->ctrl_type == DW_MCI_TYPE_ARTPEC8)
 		clksel = mci_readl(host, CLKSEL64);
 	else
 		clksel = mci_readl(host, CLKSEL);
 	clksel = SDMMC_CLKSEL_UP_SAMPLE(clksel, sample);
 	if (priv->ctrl_type == DW_MCI_TYPE_EXYNOS7 ||
-		priv->ctrl_type == DW_MCI_TYPE_EXYNOS7_SMU)
+		priv->ctrl_type == DW_MCI_TYPE_EXYNOS7_SMU ||
+		priv->ctrl_type == DW_MCI_TYPE_ARTPEC8)
 		mci_writel(host, CLKSEL64, clksel);
 	else
 		mci_writel(host, CLKSEL, clksel);
@@ -425,7 +443,8 @@ static inline u8 dw_mci_exynos_move_next_clksmpl(struct dw_mci *host)
 	u8 sample;
 
 	if (priv->ctrl_type == DW_MCI_TYPE_EXYNOS7 ||
-		priv->ctrl_type == DW_MCI_TYPE_EXYNOS7_SMU)
+		priv->ctrl_type == DW_MCI_TYPE_EXYNOS7_SMU ||
+		priv->ctrl_type == DW_MCI_TYPE_ARTPEC8)
 		clksel = mci_readl(host, CLKSEL64);
 	else
 		clksel = mci_readl(host, CLKSEL);
@@ -434,7 +453,8 @@ static inline u8 dw_mci_exynos_move_next_clksmpl(struct dw_mci *host)
 	clksel = SDMMC_CLKSEL_UP_SAMPLE(clksel, sample);
 
 	if (priv->ctrl_type == DW_MCI_TYPE_EXYNOS7 ||
-		priv->ctrl_type == DW_MCI_TYPE_EXYNOS7_SMU)
+		priv->ctrl_type == DW_MCI_TYPE_EXYNOS7_SMU ||
+		priv->ctrl_type == DW_MCI_TYPE_ARTPEC8)
 		mci_writel(host, CLKSEL64, clksel);
 	else
 		mci_writel(host, CLKSEL, clksel);
@@ -524,17 +544,65 @@ static int dw_mci_exynos_prepare_hs400_tuning(struct dw_mci *host,
 	return 0;
 }
 
+static void dw_mci_exynos_set_data_timeout(struct dw_mci *host,
+					   unsigned int timeout_ns)
+{
+	u32 clk_div, tmout;
+	u64 tmp;
+	unsigned int tmp2;
+
+	clk_div = (mci_readl(host, CLKDIV) & 0xFF) * 2;
+	if (clk_div == 0)
+		clk_div = 1;
+
+	tmp = DIV_ROUND_UP_ULL((u64)timeout_ns * host->bus_hz, NSEC_PER_SEC);
+	tmp = DIV_ROUND_UP_ULL(tmp, clk_div);
+
+	/* TMOUT[7:0] (RESPONSE_TIMEOUT) */
+	tmout = 0xFF; /* Set maximum */
+
+	/*
+	 * Extended HW timer (max = 0x6FFFFF2):
+	 * ((TMOUT[10:8] - 1) * 0xFFFFFF + TMOUT[31:11] * 8)
+	 */
+	if (!tmp || tmp > 0x6FFFFF2)
+		tmout |= (0xFFFFFF << 8);
+	else {
+		/* TMOUT[10:8] */
+		tmp2 = (((unsigned int)tmp / 0xFFFFFF) + 1) & 0x7;
+		tmout |= tmp2 << 8;
+
+		/* TMOUT[31:11] */
+		tmp = tmp - ((tmp2 - 1) * 0xFFFFFF);
+		tmout |= (tmp & 0xFFFFF8) << 8;
+	}
+
+	mci_writel(host, TMOUT, tmout);
+	dev_dbg(host->dev, "timeout_ns: %u => TMOUT[31:8]: %#08x",
+		timeout_ns, tmout >> 8);
+}
+
+static u32 dw_mci_exynos_get_drto_clks(struct dw_mci *host)
+{
+	u32 drto_clks;
+
+	drto_clks = mci_readl(host, TMOUT) >> 8;
+
+	return (((drto_clks & 0x7) - 1) * 0xFFFFFF) + ((drto_clks & 0xFFFFF8));
+}
+
 /* Common capabilities of Exynos4/Exynos5 SoC */
 static unsigned long exynos_dwmmc_caps[4] = {
-	MMC_CAP_1_8V_DDR | MMC_CAP_8_BIT_DATA | MMC_CAP_CMD23,
-	MMC_CAP_CMD23,
-	MMC_CAP_CMD23,
-	MMC_CAP_CMD23,
+	MMC_CAP_1_8V_DDR | MMC_CAP_8_BIT_DATA,
+	0,
+	0,
+	0,
 };
 
 static const struct dw_mci_drv_data exynos_drv_data = {
 	.caps			= exynos_dwmmc_caps,
 	.num_caps		= ARRAY_SIZE(exynos_dwmmc_caps),
+	.common_caps		= MMC_CAP_CMD23,
 	.init			= dw_mci_exynos_priv_init,
 	.set_ios		= dw_mci_exynos_set_ios,
 	.parse_dt		= dw_mci_exynos_parse_dt,
@@ -542,6 +610,16 @@ static const struct dw_mci_drv_data exynos_drv_data = {
 	.prepare_hs400_tuning	= dw_mci_exynos_prepare_hs400_tuning,
 };
 
+static const struct dw_mci_drv_data artpec_drv_data = {
+	.common_caps		= MMC_CAP_CMD23,
+	.init			= dw_mci_exynos_priv_init,
+	.set_ios		= dw_mci_exynos_set_ios,
+	.parse_dt		= dw_mci_exynos_parse_dt,
+	.execute_tuning		= dw_mci_exynos_execute_tuning,
+	.set_data_timeout		= dw_mci_exynos_set_data_timeout,
+	.get_drto_clks		= dw_mci_exynos_get_drto_clks,
+};
+
 static const struct of_device_id dw_mci_exynos_match[] = {
 	{ .compatible = "samsung,exynos4412-dw-mshc",
 			.data = &exynos_drv_data, },
@@ -555,6 +633,8 @@ static const struct of_device_id dw_mci_exynos_match[] = {
 			.data = &exynos_drv_data, },
 	{ .compatible = "samsung,exynos7-dw-mshc-smu",
 			.data = &exynos_drv_data, },
+	{ .compatible = "axis,artpec8-dw-mshc",
+			.data = &artpec_drv_data, },
 	{},
 };
 MODULE_DEVICE_TABLE(of, dw_mci_exynos_match);
diff --git a/drivers/mmc/host/dw_mmc-hi3798cv200.c b/drivers/mmc/host/dw_mmc-hi3798cv200.c
index 39794f93826f..e9437ef8ef19 100644
--- a/drivers/mmc/host/dw_mmc-hi3798cv200.c
+++ b/drivers/mmc/host/dw_mmc-hi3798cv200.c
@@ -23,12 +23,6 @@ struct hi3798cv200_priv {
 	struct clk *drive_clk;
 };
 
-static unsigned long dw_mci_hi3798cv200_caps[] = {
-	MMC_CAP_CMD23,
-	MMC_CAP_CMD23,
-	MMC_CAP_CMD23
-};
-
 static void dw_mci_hi3798cv200_set_ios(struct dw_mci *host, struct mmc_ios *ios)
 {
 	struct hi3798cv200_priv *priv = host->priv;
@@ -166,8 +160,7 @@ disable_sample_clk:
 }
 
 static const struct dw_mci_drv_data hi3798cv200_data = {
-	.caps = dw_mci_hi3798cv200_caps,
-	.num_caps = ARRAY_SIZE(dw_mci_hi3798cv200_caps),
+	.common_caps = MMC_CAP_CMD23,
 	.init = dw_mci_hi3798cv200_init,
 	.set_ios = dw_mci_hi3798cv200_set_ios,
 	.execute_tuning = dw_mci_hi3798cv200_execute_tuning,
diff --git a/drivers/mmc/host/dw_mmc-rockchip.c b/drivers/mmc/host/dw_mmc-rockchip.c
index d36991acd6df..95d0ec0f5f3a 100644
--- a/drivers/mmc/host/dw_mmc-rockchip.c
+++ b/drivers/mmc/host/dw_mmc-rockchip.c
@@ -300,21 +300,12 @@ static int dw_mci_rockchip_init(struct dw_mci *host)
 	return 0;
 }
 
-/* Common capabilities of RK3288 SoC */
-static unsigned long dw_mci_rk3288_dwmmc_caps[4] = {
-	MMC_CAP_CMD23,
-	MMC_CAP_CMD23,
-	MMC_CAP_CMD23,
-	MMC_CAP_CMD23,
-};
-
 static const struct dw_mci_drv_data rk2928_drv_data = {
 	.init			= dw_mci_rockchip_init,
 };
 
 static const struct dw_mci_drv_data rk3288_drv_data = {
-	.caps			= dw_mci_rk3288_dwmmc_caps,
-	.num_caps		= ARRAY_SIZE(dw_mci_rk3288_dwmmc_caps),
+	.common_caps		= MMC_CAP_CMD23,
 	.set_ios		= dw_mci_rk3288_set_ios,
 	.execute_tuning		= dw_mci_rk3288_execute_tuning,
 	.parse_dt		= dw_mci_rk3288_parse_dt,
diff --git a/drivers/mmc/host/dw_mmc.c b/drivers/mmc/host/dw_mmc.c
index d977f34f6b55..42bf8a2287ba 100644
--- a/drivers/mmc/host/dw_mmc.c
+++ b/drivers/mmc/host/dw_mmc.c
@@ -335,7 +335,8 @@ static u32 dw_mci_prep_stop_abort(struct dw_mci *host, struct mmc_command *cmd)
 	    cmdr == MMC_WRITE_BLOCK ||
 	    cmdr == MMC_WRITE_MULTIPLE_BLOCK ||
 	    cmdr == MMC_SEND_TUNING_BLOCK ||
-	    cmdr == MMC_SEND_TUNING_BLOCK_HS200) {
+	    cmdr == MMC_SEND_TUNING_BLOCK_HS200 ||
+	    cmdr == MMC_GEN_CMD) {
 		stop->opcode = MMC_STOP_TRANSMISSION;
 		stop->arg = 0;
 		stop->flags = MMC_RSP_R1B | MMC_CMD_AC;
@@ -1283,6 +1284,37 @@ static void dw_mci_setup_bus(struct dw_mci_slot *slot, bool force_clkinit)
 	mci_writel(host, CTYPE, (slot->ctype << slot->id));
 }
 
+static void dw_mci_set_data_timeout(struct dw_mci *host,
+				    unsigned int timeout_ns)
+{
+	const struct dw_mci_drv_data *drv_data = host->drv_data;
+	u32 clk_div, tmout;
+	u64 tmp;
+
+	if (drv_data && drv_data->set_data_timeout)
+		return drv_data->set_data_timeout(host, timeout_ns);
+
+	clk_div = (mci_readl(host, CLKDIV) & 0xFF) * 2;
+	if (clk_div == 0)
+		clk_div = 1;
+
+	tmp = DIV_ROUND_UP_ULL((u64)timeout_ns * host->bus_hz, NSEC_PER_SEC);
+	tmp = DIV_ROUND_UP_ULL(tmp, clk_div);
+
+	/* TMOUT[7:0] (RESPONSE_TIMEOUT) */
+	tmout = 0xFF; /* Set maximum */
+
+	/* TMOUT[31:8] (DATA_TIMEOUT) */
+	if (!tmp || tmp > 0xFFFFFF)
+		tmout |= (0xFFFFFF << 8);
+	else
+		tmout |= (tmp & 0xFFFFFF) << 8;
+
+	mci_writel(host, TMOUT, tmout);
+	dev_dbg(host->dev, "timeout_ns: %u => TMOUT[31:8]: %#08x",
+		timeout_ns, tmout >> 8);
+}
+
 static void __dw_mci_start_request(struct dw_mci *host,
 				   struct dw_mci_slot *slot,
 				   struct mmc_command *cmd)
@@ -1303,7 +1335,7 @@ static void __dw_mci_start_request(struct dw_mci *host,
 
 	data = cmd->data;
 	if (data) {
-		mci_writel(host, TMOUT, 0xFFFFFFFF);
+		dw_mci_set_data_timeout(host, data->timeout_ns);
 		mci_writel(host, BYTCNT, data->blksz*data->blocks);
 		mci_writel(host, BLKSIZ, data->blksz);
 	}
@@ -1967,12 +1999,16 @@ static int dw_mci_data_complete(struct dw_mci *host, struct mmc_data *data)
 
 static void dw_mci_set_drto(struct dw_mci *host)
 {
+	const struct dw_mci_drv_data *drv_data = host->drv_data;
 	unsigned int drto_clks;
 	unsigned int drto_div;
 	unsigned int drto_ms;
 	unsigned long irqflags;
 
-	drto_clks = mci_readl(host, TMOUT) >> 8;
+	if (drv_data && drv_data->get_drto_clks)
+		drto_clks = drv_data->get_drto_clks(host);
+	else
+		drto_clks = mci_readl(host, TMOUT) >> 8;
 	drto_div = (mci_readl(host, CLKDIV) & 0xff) * 2;
 	if (drto_div == 0)
 		drto_div = 1;
@@ -1980,6 +2016,8 @@ static void dw_mci_set_drto(struct dw_mci *host)
 	drto_ms = DIV_ROUND_UP_ULL((u64)MSEC_PER_SEC * drto_clks * drto_div,
 				   host->bus_hz);
 
+	dev_dbg(host->dev, "drto_ms: %u\n", drto_ms);
+
 	/* add a bit spare time */
 	drto_ms += 10;
 
@@ -2724,11 +2762,20 @@ static irqreturn_t dw_mci_interrupt(int irq, void *dev_id)
 		if (pending & DW_MCI_DATA_ERROR_FLAGS) {
 			spin_lock(&host->irq_lock);
 
+			if (host->quirks & DW_MMC_QUIRK_EXTENDED_TMOUT)
+				del_timer(&host->dto_timer);
+
 			/* if there is an error report DATA_ERROR */
 			mci_writel(host, RINTSTS, DW_MCI_DATA_ERROR_FLAGS);
 			host->data_status = pending;
 			smp_wmb(); /* drain writebuffer */
 			set_bit(EVENT_DATA_ERROR, &host->pending_events);
+
+			if (host->quirks & DW_MMC_QUIRK_EXTENDED_TMOUT)
+				/* In case of error, we cannot expect a DTO */
+				set_bit(EVENT_DATA_COMPLETE,
+					&host->pending_events);
+
 			tasklet_schedule(&host->tasklet);
 
 			spin_unlock(&host->irq_lock);
@@ -2828,6 +2875,9 @@ static int dw_mci_init_slot_caps(struct dw_mci_slot *slot)
 	if (host->pdata->pm_caps)
 		mmc->pm_caps = host->pdata->pm_caps;
 
+	if (drv_data)
+		mmc->caps |= drv_data->common_caps;
+
 	if (host->dev->of_node) {
 		ctrl_id = of_alias_get_id(host->dev->of_node, "mshc");
 		if (ctrl_id < 0)
diff --git a/drivers/mmc/host/dw_mmc.h b/drivers/mmc/host/dw_mmc.h
index ce05d81477d9..7f1e38621d13 100644
--- a/drivers/mmc/host/dw_mmc.h
+++ b/drivers/mmc/host/dw_mmc.h
@@ -118,6 +118,7 @@ struct dw_mci_dma_slave {
  * @part_buf: Simple buffer for partial fifo reads/writes.
  * @push_data: Pointer to FIFO push function.
  * @pull_data: Pointer to FIFO pull function.
+ * @quirks: Set of quirks that apply to specific versions of the IP.
  * @vqmmc_enabled: Status of vqmmc, should be true or false.
  * @irq_flags: The flags to be passed to request_irq.
  * @irq: The irq value to be passed to request_irq.
@@ -223,6 +224,7 @@ struct dw_mci {
 	void (*push_data)(struct dw_mci *host, void *buf, int cnt);
 	void (*pull_data)(struct dw_mci *host, void *buf, int cnt);
 
+	u32			quirks;
 	bool			vqmmc_enabled;
 	unsigned long		irq_flags; /* IRQ flags */
 	int			irq;
@@ -274,6 +276,9 @@ struct dw_mci_board {
 	struct dma_pdata *data;
 };
 
+/* Support for longer data read timeout */
+#define DW_MMC_QUIRK_EXTENDED_TMOUT            BIT(0)
+
 #define DW_MMC_240A		0x240a
 #define DW_MMC_280A		0x280a
 
@@ -550,10 +555,14 @@ struct dw_mci_slot {
  * dw_mci driver data - dw-mshc implementation specific driver data.
  * @caps: mmc subsystem specified capabilities of the controller(s).
  * @num_caps: number of capabilities specified by @caps.
+ * @common_caps: mmc subsystem specified capabilities applicable to all of
+ *	the controllers
  * @init: early implementation specific initialization.
  * @set_ios: handle bus specific extensions.
  * @parse_dt: parse implementation specific device tree properties.
  * @execute_tuning: implementation specific tuning procedure.
+ * @set_data_timeout: implementation specific timeout.
+ * @get_drto_clks: implementation specific cycle count for data read timeout.
  *
  * Provide controller implementation specific extensions. The usage of this
  * data structure is fully optional and usage of each member in this structure
@@ -562,6 +571,7 @@ struct dw_mci_slot {
 struct dw_mci_drv_data {
 	unsigned long	*caps;
 	u32		num_caps;
+	u32		common_caps;
 	int		(*init)(struct dw_mci *host);
 	void		(*set_ios)(struct dw_mci *host, struct mmc_ios *ios);
 	int		(*parse_dt)(struct dw_mci *host);
@@ -570,5 +580,8 @@ struct dw_mci_drv_data {
 						struct mmc_ios *ios);
 	int		(*switch_voltage)(struct mmc_host *mmc,
 					  struct mmc_ios *ios);
+	void		(*set_data_timeout)(struct dw_mci *host,
+					  unsigned int timeout_ns);
+	u32		(*get_drto_clks)(struct dw_mci *host);
 };
 #endif /* _DW_MMC_H_ */
diff --git a/drivers/mmc/host/jz4740_mmc.c b/drivers/mmc/host/jz4740_mmc.c
index bb612fce7ead..7693236c946f 100644
--- a/drivers/mmc/host/jz4740_mmc.c
+++ b/drivers/mmc/host/jz4740_mmc.c
@@ -217,11 +217,23 @@ static void jz4740_mmc_release_dma_channels(struct jz4740_mmc_host *host)
 		return;
 
 	dma_release_channel(host->dma_tx);
-	dma_release_channel(host->dma_rx);
+	if (host->dma_rx)
+		dma_release_channel(host->dma_rx);
 }
 
 static int jz4740_mmc_acquire_dma_channels(struct jz4740_mmc_host *host)
 {
+	struct device *dev = mmc_dev(host->mmc);
+
+	host->dma_tx = dma_request_chan(dev, "tx-rx");
+	if (!IS_ERR(host->dma_tx))
+		return 0;
+
+	if (PTR_ERR(host->dma_tx) != -ENODEV) {
+		dev_err(dev, "Failed to get dma tx-rx channel\n");
+		return PTR_ERR(host->dma_tx);
+	}
+
 	host->dma_tx = dma_request_chan(mmc_dev(host->mmc), "tx");
 	if (IS_ERR(host->dma_tx)) {
 		dev_err(mmc_dev(host->mmc), "Failed to get dma_tx channel\n");
@@ -241,7 +253,10 @@ static int jz4740_mmc_acquire_dma_channels(struct jz4740_mmc_host *host)
 static inline struct dma_chan *jz4740_mmc_get_dma_chan(struct jz4740_mmc_host *host,
 						       struct mmc_data *data)
 {
-	return (data->flags & MMC_DATA_READ) ? host->dma_rx : host->dma_tx;
+	if ((data->flags & MMC_DATA_READ) && host->dma_rx)
+		return host->dma_rx;
+	else
+		return host->dma_tx;
 }
 
 static void jz4740_mmc_dma_unmap(struct jz4740_mmc_host *host,
diff --git a/drivers/mmc/host/meson-mx-sdhc-clkc.c b/drivers/mmc/host/meson-mx-sdhc-clkc.c
index e1f29b279123..19200b7079a6 100644
--- a/drivers/mmc/host/meson-mx-sdhc-clkc.c
+++ b/drivers/mmc/host/meson-mx-sdhc-clkc.c
@@ -12,8 +12,6 @@
 
 #include "meson-mx-sdhc.h"
 
-#define MESON_SDHC_NUM_BUILTIN_CLKS	6
-
 struct meson_mx_sdhc_clkc {
 	struct clk_mux			src_sel;
 	struct clk_divider		div;
diff --git a/drivers/mmc/host/meson-mx-sdhc-mmc.c b/drivers/mmc/host/meson-mx-sdhc-mmc.c
index 8fdd0bbbfa21..28aa78aa08f3 100644
--- a/drivers/mmc/host/meson-mx-sdhc-mmc.c
+++ b/drivers/mmc/host/meson-mx-sdhc-mmc.c
@@ -854,6 +854,11 @@ static int meson_mx_sdhc_probe(struct platform_device *pdev)
 		goto err_disable_pclk;
 
 	irq = platform_get_irq(pdev, 0);
+	if (irq < 0) {
+		ret = irq;
+		goto err_disable_pclk;
+	}
+
 	ret = devm_request_threaded_irq(dev, irq, meson_mx_sdhc_irq,
 					meson_mx_sdhc_irq_thread, IRQF_ONESHOT,
 					NULL, host);
diff --git a/drivers/mmc/host/meson-mx-sdio.c b/drivers/mmc/host/meson-mx-sdio.c
index d4a48916bfb6..3a19a05ef55a 100644
--- a/drivers/mmc/host/meson-mx-sdio.c
+++ b/drivers/mmc/host/meson-mx-sdio.c
@@ -662,6 +662,11 @@ static int meson_mx_mmc_probe(struct platform_device *pdev)
 	}
 
 	irq = platform_get_irq(pdev, 0);
+	if (irq < 0) {
+		ret = irq;
+		goto error_free_mmc;
+	}
+
 	ret = devm_request_threaded_irq(host->controller_dev, irq,
 					meson_mx_mmc_irq,
 					meson_mx_mmc_irq_thread, IRQF_ONESHOT,
diff --git a/drivers/mmc/host/mmc_spi.c b/drivers/mmc/host/mmc_spi.c
index b431cdd27353..a576181e9db0 100644
--- a/drivers/mmc/host/mmc_spi.c
+++ b/drivers/mmc/host/mmc_spi.c
@@ -547,7 +547,7 @@ mmc_spi_command_send(struct mmc_spi_host *host,
 static void
 mmc_spi_setup_data_message(
 	struct mmc_spi_host	*host,
-	int			multiple,
+	bool			multiple,
 	enum dma_data_direction	direction)
 {
 	struct spi_transfer	*t;
@@ -859,14 +859,14 @@ mmc_spi_data_do(struct mmc_spi_host *host, struct mmc_command *cmd,
 	struct spi_device	*spi = host->spi;
 	struct device		*dma_dev = host->dma_dev;
 	struct spi_transfer	*t;
-	enum dma_data_direction	direction;
+	enum dma_data_direction	direction = mmc_get_dma_dir(data);
 	struct scatterlist	*sg;
 	unsigned		n_sg;
-	int			multiple = (data->blocks > 1);
+	bool			multiple = (data->blocks > 1);
+	const char		*write_or_read = (direction == DMA_TO_DEVICE) ? "write" : "read";
 	u32			clock_rate;
 	unsigned long		timeout;
 
-	direction = mmc_get_dma_dir(data);
 	mmc_spi_setup_data_message(host, multiple, direction);
 	t = &host->t;
 
@@ -921,9 +921,7 @@ mmc_spi_data_do(struct mmc_spi_host *host, struct mmc_command *cmd,
 		while (length) {
 			t->len = min(length, blk_size);
 
-			dev_dbg(&host->spi->dev, "    %s block, %d bytes\n",
-				(direction == DMA_TO_DEVICE) ? "write" : "read",
-				t->len);
+			dev_dbg(&spi->dev, "    %s block, %d bytes\n", write_or_read, t->len);
 
 			if (direction == DMA_TO_DEVICE)
 				status = mmc_spi_writeblock(host, t, timeout);
@@ -948,9 +946,7 @@ mmc_spi_data_do(struct mmc_spi_host *host, struct mmc_command *cmd,
 
 		if (status < 0) {
 			data->error = status;
-			dev_dbg(&spi->dev, "%s status %d\n",
-				(direction == DMA_TO_DEVICE) ? "write" : "read",
-				status);
+			dev_dbg(&spi->dev, "%s status %d\n", write_or_read, status);
 			break;
 		}
 	}
diff --git a/drivers/mmc/host/mmci.c b/drivers/mmc/host/mmci.c
index c9cacd4d5b22..45b8608c935c 100644
--- a/drivers/mmc/host/mmci.c
+++ b/drivers/mmc/host/mmci.c
@@ -280,7 +280,7 @@ static struct variant_data variant_stm32_sdmmc = {
 static struct variant_data variant_stm32_sdmmcv2 = {
 	.fifosize		= 16 * 4,
 	.fifohalfsize		= 8 * 4,
-	.f_max			= 208000000,
+	.f_max			= 267000000,
 	.stm32_clkdiv		= true,
 	.cmdreg_cpsm_enable	= MCI_CPSM_STM32_ENABLE,
 	.cmdreg_lrsp_crc	= MCI_CPSM_STM32_LRSP_CRC,
@@ -2435,6 +2435,11 @@ static const struct amba_id mmci_ids[] = {
 		.mask	= 0xf0ffffff,
 		.data	= &variant_stm32_sdmmcv2,
 	},
+	{
+		.id     = 0x20253180,
+		.mask	= 0xf0ffffff,
+		.data	= &variant_stm32_sdmmcv2,
+	},
 	/* Qualcomm variants */
 	{
 		.id     = 0x00051180,
diff --git a/drivers/mmc/host/mmci_stm32_sdmmc.c b/drivers/mmc/host/mmci_stm32_sdmmc.c
index a75d3dd34d18..9c13f2c31365 100644
--- a/drivers/mmc/host/mmci_stm32_sdmmc.c
+++ b/drivers/mmc/host/mmci_stm32_sdmmc.c
@@ -241,11 +241,12 @@ static void mmci_sdmmc_set_clkreg(struct mmci_host *host, unsigned int desired)
 
 	/*
 	 * SDMMC_FBCK is selected when an external Delay Block is needed
-	 * with SDR104.
+	 * with SDR104 or HS200.
 	 */
 	if (host->mmc->ios.timing >= MMC_TIMING_UHS_SDR50) {
 		clk |= MCI_STM32_CLK_BUSSPEED;
-		if (host->mmc->ios.timing == MMC_TIMING_UHS_SDR104) {
+		if (host->mmc->ios.timing == MMC_TIMING_UHS_SDR104 ||
+		    host->mmc->ios.timing == MMC_TIMING_MMC_HS200) {
 			clk &= ~MCI_STM32_CLK_SEL_MSK;
 			clk |= MCI_STM32_CLK_SELFBCK;
 		}
diff --git a/drivers/mmc/host/mtk-sd.c b/drivers/mmc/host/mtk-sd.c
index 632775217d35..65037e1d7723 100644
--- a/drivers/mmc/host/mtk-sd.c
+++ b/drivers/mmc/host/mtk-sd.c
@@ -5,6 +5,7 @@
  */
 
 #include <linux/module.h>
+#include <linux/bitops.h>
 #include <linux/clk.h>
 #include <linux/delay.h>
 #include <linux/dma-mapping.h>
@@ -98,226 +99,226 @@
 /*--------------------------------------------------------------------------*/
 
 /* MSDC_CFG mask */
-#define MSDC_CFG_MODE           (0x1 << 0)	/* RW */
-#define MSDC_CFG_CKPDN          (0x1 << 1)	/* RW */
-#define MSDC_CFG_RST            (0x1 << 2)	/* RW */
-#define MSDC_CFG_PIO            (0x1 << 3)	/* RW */
-#define MSDC_CFG_CKDRVEN        (0x1 << 4)	/* RW */
-#define MSDC_CFG_BV18SDT        (0x1 << 5)	/* RW */
-#define MSDC_CFG_BV18PSS        (0x1 << 6)	/* R  */
-#define MSDC_CFG_CKSTB          (0x1 << 7)	/* R  */
-#define MSDC_CFG_CKDIV          (0xff << 8)	/* RW */
-#define MSDC_CFG_CKMOD          (0x3 << 16)	/* RW */
-#define MSDC_CFG_HS400_CK_MODE  (0x1 << 18)	/* RW */
-#define MSDC_CFG_HS400_CK_MODE_EXTRA  (0x1 << 22)	/* RW */
-#define MSDC_CFG_CKDIV_EXTRA    (0xfff << 8)	/* RW */
-#define MSDC_CFG_CKMOD_EXTRA    (0x3 << 20)	/* RW */
+#define MSDC_CFG_MODE           BIT(0)	/* RW */
+#define MSDC_CFG_CKPDN          BIT(1)	/* RW */
+#define MSDC_CFG_RST            BIT(2)	/* RW */
+#define MSDC_CFG_PIO            BIT(3)	/* RW */
+#define MSDC_CFG_CKDRVEN        BIT(4)	/* RW */
+#define MSDC_CFG_BV18SDT        BIT(5)	/* RW */
+#define MSDC_CFG_BV18PSS        BIT(6)	/* R  */
+#define MSDC_CFG_CKSTB          BIT(7)	/* R  */
+#define MSDC_CFG_CKDIV          GENMASK(15, 8)	/* RW */
+#define MSDC_CFG_CKMOD          GENMASK(17, 16)	/* RW */
+#define MSDC_CFG_HS400_CK_MODE  BIT(18)	/* RW */
+#define MSDC_CFG_HS400_CK_MODE_EXTRA  BIT(22)	/* RW */
+#define MSDC_CFG_CKDIV_EXTRA    GENMASK(19, 8)	/* RW */
+#define MSDC_CFG_CKMOD_EXTRA    GENMASK(21, 20)	/* RW */
 
 /* MSDC_IOCON mask */
-#define MSDC_IOCON_SDR104CKS    (0x1 << 0)	/* RW */
-#define MSDC_IOCON_RSPL         (0x1 << 1)	/* RW */
-#define MSDC_IOCON_DSPL         (0x1 << 2)	/* RW */
-#define MSDC_IOCON_DDLSEL       (0x1 << 3)	/* RW */
-#define MSDC_IOCON_DDR50CKD     (0x1 << 4)	/* RW */
-#define MSDC_IOCON_DSPLSEL      (0x1 << 5)	/* RW */
-#define MSDC_IOCON_W_DSPL       (0x1 << 8)	/* RW */
-#define MSDC_IOCON_D0SPL        (0x1 << 16)	/* RW */
-#define MSDC_IOCON_D1SPL        (0x1 << 17)	/* RW */
-#define MSDC_IOCON_D2SPL        (0x1 << 18)	/* RW */
-#define MSDC_IOCON_D3SPL        (0x1 << 19)	/* RW */
-#define MSDC_IOCON_D4SPL        (0x1 << 20)	/* RW */
-#define MSDC_IOCON_D5SPL        (0x1 << 21)	/* RW */
-#define MSDC_IOCON_D6SPL        (0x1 << 22)	/* RW */
-#define MSDC_IOCON_D7SPL        (0x1 << 23)	/* RW */
-#define MSDC_IOCON_RISCSZ       (0x3 << 24)	/* RW */
+#define MSDC_IOCON_SDR104CKS    BIT(0)	/* RW */
+#define MSDC_IOCON_RSPL         BIT(1)	/* RW */
+#define MSDC_IOCON_DSPL         BIT(2)	/* RW */
+#define MSDC_IOCON_DDLSEL       BIT(3)	/* RW */
+#define MSDC_IOCON_DDR50CKD     BIT(4)	/* RW */
+#define MSDC_IOCON_DSPLSEL      BIT(5)	/* RW */
+#define MSDC_IOCON_W_DSPL       BIT(8)	/* RW */
+#define MSDC_IOCON_D0SPL        BIT(16)	/* RW */
+#define MSDC_IOCON_D1SPL        BIT(17)	/* RW */
+#define MSDC_IOCON_D2SPL        BIT(18)	/* RW */
+#define MSDC_IOCON_D3SPL        BIT(19)	/* RW */
+#define MSDC_IOCON_D4SPL        BIT(20)	/* RW */
+#define MSDC_IOCON_D5SPL        BIT(21)	/* RW */
+#define MSDC_IOCON_D6SPL        BIT(22)	/* RW */
+#define MSDC_IOCON_D7SPL        BIT(23)	/* RW */
+#define MSDC_IOCON_RISCSZ       GENMASK(25, 24)	/* RW */
 
 /* MSDC_PS mask */
-#define MSDC_PS_CDEN            (0x1 << 0)	/* RW */
-#define MSDC_PS_CDSTS           (0x1 << 1)	/* R  */
-#define MSDC_PS_CDDEBOUNCE      (0xf << 12)	/* RW */
-#define MSDC_PS_DAT             (0xff << 16)	/* R  */
-#define MSDC_PS_DATA1           (0x1 << 17)	/* R  */
-#define MSDC_PS_CMD             (0x1 << 24)	/* R  */
-#define MSDC_PS_WP              (0x1 << 31)	/* R  */
+#define MSDC_PS_CDEN            BIT(0)	/* RW */
+#define MSDC_PS_CDSTS           BIT(1)	/* R  */
+#define MSDC_PS_CDDEBOUNCE      GENMASK(15, 12)	/* RW */
+#define MSDC_PS_DAT             GENMASK(23, 16)	/* R  */
+#define MSDC_PS_DATA1           BIT(17)	/* R  */
+#define MSDC_PS_CMD             BIT(24)	/* R  */
+#define MSDC_PS_WP              BIT(31)	/* R  */
 
 /* MSDC_INT mask */
-#define MSDC_INT_MMCIRQ         (0x1 << 0)	/* W1C */
-#define MSDC_INT_CDSC           (0x1 << 1)	/* W1C */
-#define MSDC_INT_ACMDRDY        (0x1 << 3)	/* W1C */
-#define MSDC_INT_ACMDTMO        (0x1 << 4)	/* W1C */
-#define MSDC_INT_ACMDCRCERR     (0x1 << 5)	/* W1C */
-#define MSDC_INT_DMAQ_EMPTY     (0x1 << 6)	/* W1C */
-#define MSDC_INT_SDIOIRQ        (0x1 << 7)	/* W1C */
-#define MSDC_INT_CMDRDY         (0x1 << 8)	/* W1C */
-#define MSDC_INT_CMDTMO         (0x1 << 9)	/* W1C */
-#define MSDC_INT_RSPCRCERR      (0x1 << 10)	/* W1C */
-#define MSDC_INT_CSTA           (0x1 << 11)	/* R */
-#define MSDC_INT_XFER_COMPL     (0x1 << 12)	/* W1C */
-#define MSDC_INT_DXFER_DONE     (0x1 << 13)	/* W1C */
-#define MSDC_INT_DATTMO         (0x1 << 14)	/* W1C */
-#define MSDC_INT_DATCRCERR      (0x1 << 15)	/* W1C */
-#define MSDC_INT_ACMD19_DONE    (0x1 << 16)	/* W1C */
-#define MSDC_INT_DMA_BDCSERR    (0x1 << 17)	/* W1C */
-#define MSDC_INT_DMA_GPDCSERR   (0x1 << 18)	/* W1C */
-#define MSDC_INT_DMA_PROTECT    (0x1 << 19)	/* W1C */
-#define MSDC_INT_CMDQ           (0x1 << 28)	/* W1C */
+#define MSDC_INT_MMCIRQ         BIT(0)	/* W1C */
+#define MSDC_INT_CDSC           BIT(1)	/* W1C */
+#define MSDC_INT_ACMDRDY        BIT(3)	/* W1C */
+#define MSDC_INT_ACMDTMO        BIT(4)	/* W1C */
+#define MSDC_INT_ACMDCRCERR     BIT(5)	/* W1C */
+#define MSDC_INT_DMAQ_EMPTY     BIT(6)	/* W1C */
+#define MSDC_INT_SDIOIRQ        BIT(7)	/* W1C */
+#define MSDC_INT_CMDRDY         BIT(8)	/* W1C */
+#define MSDC_INT_CMDTMO         BIT(9)	/* W1C */
+#define MSDC_INT_RSPCRCERR      BIT(10)	/* W1C */
+#define MSDC_INT_CSTA           BIT(11)	/* R */
+#define MSDC_INT_XFER_COMPL     BIT(12)	/* W1C */
+#define MSDC_INT_DXFER_DONE     BIT(13)	/* W1C */
+#define MSDC_INT_DATTMO         BIT(14)	/* W1C */
+#define MSDC_INT_DATCRCERR      BIT(15)	/* W1C */
+#define MSDC_INT_ACMD19_DONE    BIT(16)	/* W1C */
+#define MSDC_INT_DMA_BDCSERR    BIT(17)	/* W1C */
+#define MSDC_INT_DMA_GPDCSERR   BIT(18)	/* W1C */
+#define MSDC_INT_DMA_PROTECT    BIT(19)	/* W1C */
+#define MSDC_INT_CMDQ           BIT(28)	/* W1C */
 
 /* MSDC_INTEN mask */
-#define MSDC_INTEN_MMCIRQ       (0x1 << 0)	/* RW */
-#define MSDC_INTEN_CDSC         (0x1 << 1)	/* RW */
-#define MSDC_INTEN_ACMDRDY      (0x1 << 3)	/* RW */
-#define MSDC_INTEN_ACMDTMO      (0x1 << 4)	/* RW */
-#define MSDC_INTEN_ACMDCRCERR   (0x1 << 5)	/* RW */
-#define MSDC_INTEN_DMAQ_EMPTY   (0x1 << 6)	/* RW */
-#define MSDC_INTEN_SDIOIRQ      (0x1 << 7)	/* RW */
-#define MSDC_INTEN_CMDRDY       (0x1 << 8)	/* RW */
-#define MSDC_INTEN_CMDTMO       (0x1 << 9)	/* RW */
-#define MSDC_INTEN_RSPCRCERR    (0x1 << 10)	/* RW */
-#define MSDC_INTEN_CSTA         (0x1 << 11)	/* RW */
-#define MSDC_INTEN_XFER_COMPL   (0x1 << 12)	/* RW */
-#define MSDC_INTEN_DXFER_DONE   (0x1 << 13)	/* RW */
-#define MSDC_INTEN_DATTMO       (0x1 << 14)	/* RW */
-#define MSDC_INTEN_DATCRCERR    (0x1 << 15)	/* RW */
-#define MSDC_INTEN_ACMD19_DONE  (0x1 << 16)	/* RW */
-#define MSDC_INTEN_DMA_BDCSERR  (0x1 << 17)	/* RW */
-#define MSDC_INTEN_DMA_GPDCSERR (0x1 << 18)	/* RW */
-#define MSDC_INTEN_DMA_PROTECT  (0x1 << 19)	/* RW */
+#define MSDC_INTEN_MMCIRQ       BIT(0)	/* RW */
+#define MSDC_INTEN_CDSC         BIT(1)	/* RW */
+#define MSDC_INTEN_ACMDRDY      BIT(3)	/* RW */
+#define MSDC_INTEN_ACMDTMO      BIT(4)	/* RW */
+#define MSDC_INTEN_ACMDCRCERR   BIT(5)	/* RW */
+#define MSDC_INTEN_DMAQ_EMPTY   BIT(6)	/* RW */
+#define MSDC_INTEN_SDIOIRQ      BIT(7)	/* RW */
+#define MSDC_INTEN_CMDRDY       BIT(8)	/* RW */
+#define MSDC_INTEN_CMDTMO       BIT(9)	/* RW */
+#define MSDC_INTEN_RSPCRCERR    BIT(10)	/* RW */
+#define MSDC_INTEN_CSTA         BIT(11)	/* RW */
+#define MSDC_INTEN_XFER_COMPL   BIT(12)	/* RW */
+#define MSDC_INTEN_DXFER_DONE   BIT(13)	/* RW */
+#define MSDC_INTEN_DATTMO       BIT(14)	/* RW */
+#define MSDC_INTEN_DATCRCERR    BIT(15)	/* RW */
+#define MSDC_INTEN_ACMD19_DONE  BIT(16)	/* RW */
+#define MSDC_INTEN_DMA_BDCSERR  BIT(17)	/* RW */
+#define MSDC_INTEN_DMA_GPDCSERR BIT(18)	/* RW */
+#define MSDC_INTEN_DMA_PROTECT  BIT(19)	/* RW */
 
 /* MSDC_FIFOCS mask */
-#define MSDC_FIFOCS_RXCNT       (0xff << 0)	/* R */
-#define MSDC_FIFOCS_TXCNT       (0xff << 16)	/* R */
-#define MSDC_FIFOCS_CLR         (0x1 << 31)	/* RW */
+#define MSDC_FIFOCS_RXCNT       GENMASK(7, 0)	/* R */
+#define MSDC_FIFOCS_TXCNT       GENMASK(23, 16)	/* R */
+#define MSDC_FIFOCS_CLR         BIT(31)	/* RW */
 
 /* SDC_CFG mask */
-#define SDC_CFG_SDIOINTWKUP     (0x1 << 0)	/* RW */
-#define SDC_CFG_INSWKUP         (0x1 << 1)	/* RW */
-#define SDC_CFG_WRDTOC          (0x1fff  << 2)  /* RW */
-#define SDC_CFG_BUSWIDTH        (0x3 << 16)	/* RW */
-#define SDC_CFG_SDIO            (0x1 << 19)	/* RW */
-#define SDC_CFG_SDIOIDE         (0x1 << 20)	/* RW */
-#define SDC_CFG_INTATGAP        (0x1 << 21)	/* RW */
-#define SDC_CFG_DTOC            (0xff << 24)	/* RW */
+#define SDC_CFG_SDIOINTWKUP     BIT(0)	/* RW */
+#define SDC_CFG_INSWKUP         BIT(1)	/* RW */
+#define SDC_CFG_WRDTOC          GENMASK(14, 2)  /* RW */
+#define SDC_CFG_BUSWIDTH        GENMASK(17, 16)	/* RW */
+#define SDC_CFG_SDIO            BIT(19)	/* RW */
+#define SDC_CFG_SDIOIDE         BIT(20)	/* RW */
+#define SDC_CFG_INTATGAP        BIT(21)	/* RW */
+#define SDC_CFG_DTOC            GENMASK(31, 24)	/* RW */
 
 /* SDC_STS mask */
-#define SDC_STS_SDCBUSY         (0x1 << 0)	/* RW */
-#define SDC_STS_CMDBUSY         (0x1 << 1)	/* RW */
-#define SDC_STS_SWR_COMPL       (0x1 << 31)	/* RW */
+#define SDC_STS_SDCBUSY         BIT(0)	/* RW */
+#define SDC_STS_CMDBUSY         BIT(1)	/* RW */
+#define SDC_STS_SWR_COMPL       BIT(31)	/* RW */
 
-#define SDC_DAT1_IRQ_TRIGGER	(0x1 << 19)	/* RW */
+#define SDC_DAT1_IRQ_TRIGGER	BIT(19)	/* RW */
 /* SDC_ADV_CFG0 mask */
-#define SDC_RX_ENHANCE_EN	(0x1 << 20)	/* RW */
+#define SDC_RX_ENHANCE_EN	BIT(20)	/* RW */
 
 /* DMA_SA_H4BIT mask */
-#define DMA_ADDR_HIGH_4BIT      (0xf << 0)      /* RW */
+#define DMA_ADDR_HIGH_4BIT      GENMASK(3, 0)	/* RW */
 
 /* MSDC_DMA_CTRL mask */
-#define MSDC_DMA_CTRL_START     (0x1 << 0)	/* W */
-#define MSDC_DMA_CTRL_STOP      (0x1 << 1)	/* W */
-#define MSDC_DMA_CTRL_RESUME    (0x1 << 2)	/* W */
-#define MSDC_DMA_CTRL_MODE      (0x1 << 8)	/* RW */
-#define MSDC_DMA_CTRL_LASTBUF   (0x1 << 10)	/* RW */
-#define MSDC_DMA_CTRL_BRUSTSZ   (0x7 << 12)	/* RW */
+#define MSDC_DMA_CTRL_START     BIT(0)	/* W */
+#define MSDC_DMA_CTRL_STOP      BIT(1)	/* W */
+#define MSDC_DMA_CTRL_RESUME    BIT(2)	/* W */
+#define MSDC_DMA_CTRL_MODE      BIT(8)	/* RW */
+#define MSDC_DMA_CTRL_LASTBUF   BIT(10)	/* RW */
+#define MSDC_DMA_CTRL_BRUSTSZ   GENMASK(14, 12)	/* RW */
 
 /* MSDC_DMA_CFG mask */
-#define MSDC_DMA_CFG_STS        (0x1 << 0)	/* R */
-#define MSDC_DMA_CFG_DECSEN     (0x1 << 1)	/* RW */
-#define MSDC_DMA_CFG_AHBHPROT2  (0x2 << 8)	/* RW */
-#define MSDC_DMA_CFG_ACTIVEEN   (0x2 << 12)	/* RW */
-#define MSDC_DMA_CFG_CS12B16B   (0x1 << 16)	/* RW */
+#define MSDC_DMA_CFG_STS        BIT(0)	/* R */
+#define MSDC_DMA_CFG_DECSEN     BIT(1)	/* RW */
+#define MSDC_DMA_CFG_AHBHPROT2  BIT(9)	/* RW */
+#define MSDC_DMA_CFG_ACTIVEEN   BIT(13)	/* RW */
+#define MSDC_DMA_CFG_CS12B16B   BIT(16)	/* RW */
 
 /* MSDC_PATCH_BIT mask */
-#define MSDC_PATCH_BIT_ODDSUPP    (0x1 <<  1)	/* RW */
-#define MSDC_INT_DAT_LATCH_CK_SEL (0x7 <<  7)
-#define MSDC_CKGEN_MSDC_DLY_SEL   (0x1f << 10)
-#define MSDC_PATCH_BIT_IODSSEL    (0x1 << 16)	/* RW */
-#define MSDC_PATCH_BIT_IOINTSEL   (0x1 << 17)	/* RW */
-#define MSDC_PATCH_BIT_BUSYDLY    (0xf << 18)	/* RW */
-#define MSDC_PATCH_BIT_WDOD       (0xf << 22)	/* RW */
-#define MSDC_PATCH_BIT_IDRTSEL    (0x1 << 26)	/* RW */
-#define MSDC_PATCH_BIT_CMDFSEL    (0x1 << 27)	/* RW */
-#define MSDC_PATCH_BIT_INTDLSEL   (0x1 << 28)	/* RW */
-#define MSDC_PATCH_BIT_SPCPUSH    (0x1 << 29)	/* RW */
-#define MSDC_PATCH_BIT_DECRCTMO   (0x1 << 30)	/* RW */
-
-#define MSDC_PATCH_BIT1_CMDTA     (0x7 << 3)    /* RW */
-#define MSDC_PB1_BUSY_CHECK_SEL   (0x1 << 7)    /* RW */
-#define MSDC_PATCH_BIT1_STOP_DLY  (0xf << 8)    /* RW */
-
-#define MSDC_PATCH_BIT2_CFGRESP   (0x1 << 15)   /* RW */
-#define MSDC_PATCH_BIT2_CFGCRCSTS (0x1 << 28)   /* RW */
-#define MSDC_PB2_SUPPORT_64G      (0x1 << 1)    /* RW */
-#define MSDC_PB2_RESPWAIT         (0x3 << 2)    /* RW */
-#define MSDC_PB2_RESPSTSENSEL     (0x7 << 16)   /* RW */
-#define MSDC_PB2_CRCSTSENSEL      (0x7 << 29)   /* RW */
-
-#define MSDC_PAD_TUNE_DATWRDLY	  (0x1f <<  0)	/* RW */
-#define MSDC_PAD_TUNE_DATRRDLY	  (0x1f <<  8)	/* RW */
-#define MSDC_PAD_TUNE_CMDRDLY	  (0x1f << 16)  /* RW */
-#define MSDC_PAD_TUNE_CMDRRDLY	  (0x1f << 22)	/* RW */
-#define MSDC_PAD_TUNE_CLKTDLY	  (0x1f << 27)  /* RW */
-#define MSDC_PAD_TUNE_RXDLYSEL	  (0x1 << 15)   /* RW */
-#define MSDC_PAD_TUNE_RD_SEL	  (0x1 << 13)   /* RW */
-#define MSDC_PAD_TUNE_CMD_SEL	  (0x1 << 21)   /* RW */
-
-#define PAD_DS_TUNE_DLY_SEL       (0x1 << 0)	/* RW */
-#define PAD_DS_TUNE_DLY1	  (0x1f << 2)   /* RW */
-#define PAD_DS_TUNE_DLY2	  (0x1f << 7)   /* RW */
-#define PAD_DS_TUNE_DLY3	  (0x1f << 12)  /* RW */
-
-#define PAD_CMD_TUNE_RX_DLY3	  (0x1f << 1)  /* RW */
+#define MSDC_PATCH_BIT_ODDSUPP    BIT(1)	/* RW */
+#define MSDC_INT_DAT_LATCH_CK_SEL GENMASK(9, 7)
+#define MSDC_CKGEN_MSDC_DLY_SEL   GENMASK(14, 10)
+#define MSDC_PATCH_BIT_IODSSEL    BIT(16)	/* RW */
+#define MSDC_PATCH_BIT_IOINTSEL   BIT(17)	/* RW */
+#define MSDC_PATCH_BIT_BUSYDLY    GENMASK(21, 18)	/* RW */
+#define MSDC_PATCH_BIT_WDOD       GENMASK(25, 22)	/* RW */
+#define MSDC_PATCH_BIT_IDRTSEL    BIT(26)	/* RW */
+#define MSDC_PATCH_BIT_CMDFSEL    BIT(27)	/* RW */
+#define MSDC_PATCH_BIT_INTDLSEL   BIT(28)	/* RW */
+#define MSDC_PATCH_BIT_SPCPUSH    BIT(29)	/* RW */
+#define MSDC_PATCH_BIT_DECRCTMO   BIT(30)	/* RW */
+
+#define MSDC_PATCH_BIT1_CMDTA     GENMASK(5, 3)    /* RW */
+#define MSDC_PB1_BUSY_CHECK_SEL   BIT(7)    /* RW */
+#define MSDC_PATCH_BIT1_STOP_DLY  GENMASK(11, 8)    /* RW */
+
+#define MSDC_PATCH_BIT2_CFGRESP   BIT(15)   /* RW */
+#define MSDC_PATCH_BIT2_CFGCRCSTS BIT(28)   /* RW */
+#define MSDC_PB2_SUPPORT_64G      BIT(1)    /* RW */
+#define MSDC_PB2_RESPWAIT         GENMASK(3, 2)   /* RW */
+#define MSDC_PB2_RESPSTSENSEL     GENMASK(18, 16) /* RW */
+#define MSDC_PB2_CRCSTSENSEL      GENMASK(31, 29) /* RW */
+
+#define MSDC_PAD_TUNE_DATWRDLY	  GENMASK(4, 0)		/* RW */
+#define MSDC_PAD_TUNE_DATRRDLY	  GENMASK(12, 8)	/* RW */
+#define MSDC_PAD_TUNE_CMDRDLY	  GENMASK(20, 16)	/* RW */
+#define MSDC_PAD_TUNE_CMDRRDLY	  GENMASK(26, 22)	/* RW */
+#define MSDC_PAD_TUNE_CLKTDLY	  GENMASK(31, 27)	/* RW */
+#define MSDC_PAD_TUNE_RXDLYSEL	  BIT(15)   /* RW */
+#define MSDC_PAD_TUNE_RD_SEL	  BIT(13)   /* RW */
+#define MSDC_PAD_TUNE_CMD_SEL	  BIT(21)   /* RW */
+
+#define PAD_DS_TUNE_DLY_SEL       BIT(0)	  /* RW */
+#define PAD_DS_TUNE_DLY1	  GENMASK(6, 2)   /* RW */
+#define PAD_DS_TUNE_DLY2	  GENMASK(11, 7)  /* RW */
+#define PAD_DS_TUNE_DLY3	  GENMASK(16, 12) /* RW */
+
+#define PAD_CMD_TUNE_RX_DLY3	  GENMASK(5, 1)   /* RW */
 
 /* EMMC51_CFG0 mask */
-#define CMDQ_RDAT_CNT		  (0x3ff << 12)	/* RW */
+#define CMDQ_RDAT_CNT		  GENMASK(21, 12) /* RW */
 
-#define EMMC50_CFG_PADCMD_LATCHCK (0x1 << 0)   /* RW */
-#define EMMC50_CFG_CRCSTS_EDGE    (0x1 << 3)   /* RW */
-#define EMMC50_CFG_CFCSTS_SEL     (0x1 << 4)   /* RW */
-#define EMMC50_CFG_CMD_RESP_SEL   (0x1 << 9)   /* RW */
+#define EMMC50_CFG_PADCMD_LATCHCK BIT(0)   /* RW */
+#define EMMC50_CFG_CRCSTS_EDGE    BIT(3)   /* RW */
+#define EMMC50_CFG_CFCSTS_SEL     BIT(4)   /* RW */
+#define EMMC50_CFG_CMD_RESP_SEL   BIT(9)   /* RW */
 
 /* EMMC50_CFG1 mask */
-#define EMMC50_CFG1_DS_CFG        (0x1 << 28)  /* RW */
+#define EMMC50_CFG1_DS_CFG        BIT(28)  /* RW */
 
-#define EMMC50_CFG3_OUTS_WR       (0x1f << 0)  /* RW */
+#define EMMC50_CFG3_OUTS_WR       GENMASK(4, 0)  /* RW */
 
-#define SDC_FIFO_CFG_WRVALIDSEL   (0x1 << 24)  /* RW */
-#define SDC_FIFO_CFG_RDVALIDSEL   (0x1 << 25)  /* RW */
+#define SDC_FIFO_CFG_WRVALIDSEL   BIT(24)  /* RW */
+#define SDC_FIFO_CFG_RDVALIDSEL   BIT(25)  /* RW */
 
 /* CQHCI_SETTING */
-#define CQHCI_RD_CMD_WND_SEL	  (0x1 << 14) /* RW */
-#define CQHCI_WR_CMD_WND_SEL	  (0x1 << 15) /* RW */
+#define CQHCI_RD_CMD_WND_SEL	  BIT(14) /* RW */
+#define CQHCI_WR_CMD_WND_SEL	  BIT(15) /* RW */
 
 /* EMMC_TOP_CONTROL mask */
-#define PAD_RXDLY_SEL           (0x1 << 0)      /* RW */
-#define DELAY_EN                (0x1 << 1)      /* RW */
-#define PAD_DAT_RD_RXDLY2       (0x1f << 2)     /* RW */
-#define PAD_DAT_RD_RXDLY        (0x1f << 7)     /* RW */
-#define PAD_DAT_RD_RXDLY2_SEL   (0x1 << 12)     /* RW */
-#define PAD_DAT_RD_RXDLY_SEL    (0x1 << 13)     /* RW */
-#define DATA_K_VALUE_SEL        (0x1 << 14)     /* RW */
-#define SDC_RX_ENH_EN           (0x1 << 15)     /* TW */
+#define PAD_RXDLY_SEL           BIT(0)      /* RW */
+#define DELAY_EN                BIT(1)      /* RW */
+#define PAD_DAT_RD_RXDLY2       GENMASK(6, 2)     /* RW */
+#define PAD_DAT_RD_RXDLY        GENMASK(11, 7)    /* RW */
+#define PAD_DAT_RD_RXDLY2_SEL   BIT(12)     /* RW */
+#define PAD_DAT_RD_RXDLY_SEL    BIT(13)     /* RW */
+#define DATA_K_VALUE_SEL        BIT(14)     /* RW */
+#define SDC_RX_ENH_EN           BIT(15)     /* TW */
 
 /* EMMC_TOP_CMD mask */
-#define PAD_CMD_RXDLY2          (0x1f << 0)     /* RW */
-#define PAD_CMD_RXDLY           (0x1f << 5)     /* RW */
-#define PAD_CMD_RD_RXDLY2_SEL   (0x1 << 10)     /* RW */
-#define PAD_CMD_RD_RXDLY_SEL    (0x1 << 11)     /* RW */
-#define PAD_CMD_TX_DLY          (0x1f << 12)    /* RW */
+#define PAD_CMD_RXDLY2          GENMASK(4, 0)	/* RW */
+#define PAD_CMD_RXDLY           GENMASK(9, 5)	/* RW */
+#define PAD_CMD_RD_RXDLY2_SEL   BIT(10)		/* RW */
+#define PAD_CMD_RD_RXDLY_SEL    BIT(11)		/* RW */
+#define PAD_CMD_TX_DLY          GENMASK(16, 12)	/* RW */
 
 /* EMMC50_PAD_DS_TUNE mask */
-#define PAD_DS_DLY_SEL		(0x1 << 16)	/* RW */
-#define PAD_DS_DLY1		(0x1f << 10)	/* RW */
-#define PAD_DS_DLY3		(0x1f << 0)	/* RW */
+#define PAD_DS_DLY_SEL		BIT(16)	/* RW */
+#define PAD_DS_DLY1		GENMASK(14, 10)	/* RW */
+#define PAD_DS_DLY3		GENMASK(4, 0)	/* RW */
 
-#define REQ_CMD_EIO  (0x1 << 0)
-#define REQ_CMD_TMO  (0x1 << 1)
-#define REQ_DAT_ERR  (0x1 << 2)
-#define REQ_STOP_EIO (0x1 << 3)
-#define REQ_STOP_TMO (0x1 << 4)
-#define REQ_CMD_BUSY (0x1 << 5)
+#define REQ_CMD_EIO  BIT(0)
+#define REQ_CMD_TMO  BIT(1)
+#define REQ_DAT_ERR  BIT(2)
+#define REQ_STOP_EIO BIT(3)
+#define REQ_STOP_TMO BIT(4)
+#define REQ_CMD_BUSY BIT(5)
 
-#define MSDC_PREPARE_FLAG (0x1 << 0)
-#define MSDC_ASYNC_FLAG (0x1 << 1)
-#define MSDC_MMAP_FLAG (0x1 << 2)
+#define MSDC_PREPARE_FLAG BIT(0)
+#define MSDC_ASYNC_FLAG BIT(1)
+#define MSDC_MMAP_FLAG BIT(2)
 
 #define MTK_MMC_AUTOSUSPEND_DELAY	50
 #define CMD_TIMEOUT         (HZ/10 * 5)	/* 100ms x5 */
@@ -331,17 +332,17 @@
 /*--------------------------------------------------------------------------*/
 struct mt_gpdma_desc {
 	u32 gpd_info;
-#define GPDMA_DESC_HWO		(0x1 << 0)
-#define GPDMA_DESC_BDP		(0x1 << 1)
-#define GPDMA_DESC_CHECKSUM	(0xff << 8) /* bit8 ~ bit15 */
-#define GPDMA_DESC_INT		(0x1 << 16)
-#define GPDMA_DESC_NEXT_H4	(0xf << 24)
-#define GPDMA_DESC_PTR_H4	(0xf << 28)
+#define GPDMA_DESC_HWO		BIT(0)
+#define GPDMA_DESC_BDP		BIT(1)
+#define GPDMA_DESC_CHECKSUM	GENMASK(15, 8)
+#define GPDMA_DESC_INT		BIT(16)
+#define GPDMA_DESC_NEXT_H4	GENMASK(27, 24)
+#define GPDMA_DESC_PTR_H4	GENMASK(31, 28)
 	u32 next;
 	u32 ptr;
 	u32 gpd_data_len;
-#define GPDMA_DESC_BUFLEN	(0xffff) /* bit0 ~ bit15 */
-#define GPDMA_DESC_EXTLEN	(0xff << 16) /* bit16 ~ bit23 */
+#define GPDMA_DESC_BUFLEN	GENMASK(15, 0)
+#define GPDMA_DESC_EXTLEN	GENMASK(23, 16)
 	u32 arg;
 	u32 blknum;
 	u32 cmd;
@@ -349,17 +350,17 @@ struct mt_gpdma_desc {
 
 struct mt_bdma_desc {
 	u32 bd_info;
-#define BDMA_DESC_EOL		(0x1 << 0)
-#define BDMA_DESC_CHECKSUM	(0xff << 8) /* bit8 ~ bit15 */
-#define BDMA_DESC_BLKPAD	(0x1 << 17)
-#define BDMA_DESC_DWPAD		(0x1 << 18)
-#define BDMA_DESC_NEXT_H4	(0xf << 24)
-#define BDMA_DESC_PTR_H4	(0xf << 28)
+#define BDMA_DESC_EOL		BIT(0)
+#define BDMA_DESC_CHECKSUM	GENMASK(15, 8)
+#define BDMA_DESC_BLKPAD	BIT(17)
+#define BDMA_DESC_DWPAD		BIT(18)
+#define BDMA_DESC_NEXT_H4	GENMASK(27, 24)
+#define BDMA_DESC_PTR_H4	GENMASK(31, 28)
 	u32 next;
 	u32 ptr;
 	u32 bd_data_len;
-#define BDMA_DESC_BUFLEN	(0xffff) /* bit0 ~ bit15 */
-#define BDMA_DESC_BUFLEN_EXT	(0xffffff) /* bit0 ~ bit23 */
+#define BDMA_DESC_BUFLEN	GENMASK(15, 0)
+#define BDMA_DESC_BUFLEN_EXT	GENMASK(23, 0)
 };
 
 struct msdc_dma {
@@ -636,12 +637,11 @@ static void msdc_reset_hw(struct msdc_host *host)
 	u32 val;
 
 	sdr_set_bits(host->base + MSDC_CFG, MSDC_CFG_RST);
-	while (readl(host->base + MSDC_CFG) & MSDC_CFG_RST)
-		cpu_relax();
+	readl_poll_timeout(host->base + MSDC_CFG, val, !(val & MSDC_CFG_RST), 0, 0);
 
 	sdr_set_bits(host->base + MSDC_FIFOCS, MSDC_FIFOCS_CLR);
-	while (readl(host->base + MSDC_FIFOCS) & MSDC_FIFOCS_CLR)
-		cpu_relax();
+	readl_poll_timeout(host->base + MSDC_FIFOCS, val,
+			   !(val & MSDC_FIFOCS_CLR), 0, 0);
 
 	val = readl(host->base + MSDC_INT);
 	writel(val, host->base + MSDC_INT);
@@ -725,7 +725,7 @@ static inline void msdc_dma_setup(struct msdc_host *host, struct msdc_dma *dma,
 	sdr_set_field(host->base + MSDC_DMA_CFG, MSDC_DMA_CFG_DECSEN, 1);
 	dma_ctrl = readl_relaxed(host->base + MSDC_DMA_CTRL);
 	dma_ctrl &= ~(MSDC_DMA_CTRL_BRUSTSZ | MSDC_DMA_CTRL_MODE);
-	dma_ctrl |= (MSDC_BURST_64B << 12 | 1 << 8);
+	dma_ctrl |= (MSDC_BURST_64B << 12 | BIT(8));
 	writel_relaxed(dma_ctrl, host->base + MSDC_DMA_CTRL);
 	if (host->dev_comp->support_64g)
 		sdr_set_field(host->base + DMA_SA_H4BIT, DMA_ADDR_HIGH_4BIT,
@@ -769,7 +769,7 @@ static u64 msdc_timeout_cal(struct msdc_host *host, u64 ns, u64 clks)
 		do_div(timeout, clk_ns);
 		timeout += clks;
 		/* in 1048576 sclk cycle unit */
-		timeout = DIV_ROUND_UP(timeout, (0x1 << 20));
+		timeout = DIV_ROUND_UP(timeout, BIT(20));
 		if (host->dev_comp->clk_div_bits == 8)
 			sdr_get_field(host->base + MSDC_CFG,
 				      MSDC_CFG_CKMOD, &mode);
@@ -814,8 +814,9 @@ static void msdc_gate_clock(struct msdc_host *host)
 	clk_disable_unprepare(host->h_clk);
 }
 
-static void msdc_ungate_clock(struct msdc_host *host)
+static int msdc_ungate_clock(struct msdc_host *host)
 {
+	u32 val;
 	int ret;
 
 	clk_prepare_enable(host->h_clk);
@@ -825,11 +826,11 @@ static void msdc_ungate_clock(struct msdc_host *host)
 	ret = clk_bulk_prepare_enable(MSDC_NR_CLOCKS, host->bulk_clks);
 	if (ret) {
 		dev_err(host->dev, "Cannot enable pclk/axi/ahb clock gates\n");
-		return;
+		return ret;
 	}
 
-	while (!(readl(host->base + MSDC_CFG) & MSDC_CFG_CKSTB))
-		cpu_relax();
+	return readl_poll_timeout(host->base + MSDC_CFG, val,
+				  (val & MSDC_CFG_CKSTB), 1, 20000);
 }
 
 static void msdc_set_mclk(struct msdc_host *host, unsigned char timing, u32 hz)
@@ -840,6 +841,7 @@ static void msdc_set_mclk(struct msdc_host *host, unsigned char timing, u32 hz)
 	u32 div;
 	u32 sclk;
 	u32 tune_reg = host->dev_comp->pad_tune_reg;
+	u32 val;
 
 	if (!hz) {
 		dev_dbg(host->dev, "set mclk to 0\n");
@@ -899,14 +901,8 @@ static void msdc_set_mclk(struct msdc_host *host, unsigned char timing, u32 hz)
 		}
 	}
 	sdr_clr_bits(host->base + MSDC_CFG, MSDC_CFG_CKPDN);
-	/*
-	 * As src_clk/HCLK use the same bit to gate/ungate,
-	 * So if want to only gate src_clk, need gate its parent(mux).
-	 */
-	if (host->src_clk_cg)
-		clk_disable_unprepare(host->src_clk_cg);
-	else
-		clk_disable_unprepare(clk_get_parent(host->src_clk));
+
+	clk_disable_unprepare(host->src_clk_cg);
 	if (host->dev_comp->clk_div_bits == 8)
 		sdr_set_field(host->base + MSDC_CFG,
 			      MSDC_CFG_CKMOD | MSDC_CFG_CKDIV,
@@ -915,13 +911,9 @@ static void msdc_set_mclk(struct msdc_host *host, unsigned char timing, u32 hz)
 		sdr_set_field(host->base + MSDC_CFG,
 			      MSDC_CFG_CKMOD_EXTRA | MSDC_CFG_CKDIV_EXTRA,
 			      (mode << 12) | div);
-	if (host->src_clk_cg)
-		clk_prepare_enable(host->src_clk_cg);
-	else
-		clk_prepare_enable(clk_get_parent(host->src_clk));
 
-	while (!(readl(host->base + MSDC_CFG) & MSDC_CFG_CKSTB))
-		cpu_relax();
+	clk_prepare_enable(host->src_clk_cg);
+	readl_poll_timeout(host->base + MSDC_CFG, val, (val & MSDC_CFG_CKSTB), 0, 0);
 	sdr_set_bits(host->base + MSDC_CFG, MSDC_CFG_CKPDN);
 	mmc->actual_clock = sclk;
 	host->mclk = hz;
@@ -1013,15 +1005,15 @@ static inline u32 msdc_cmd_prepare_raw_cmd(struct msdc_host *host,
 
 	if ((opcode == SD_IO_RW_DIRECT && cmd->flags == (unsigned int) -1) ||
 	    opcode == MMC_STOP_TRANSMISSION)
-		rawcmd |= (0x1 << 14);
+		rawcmd |= BIT(14);
 	else if (opcode == SD_SWITCH_VOLTAGE)
-		rawcmd |= (0x1 << 30);
+		rawcmd |= BIT(30);
 	else if (opcode == SD_APP_SEND_SCR ||
 		 opcode == SD_APP_SEND_NUM_WR_BLKS ||
 		 (opcode == SD_SWITCH && mmc_cmd_type(cmd) == MMC_CMD_ADTC) ||
 		 (opcode == SD_APP_SD_STATUS && mmc_cmd_type(cmd) == MMC_CMD_ADTC) ||
 		 (opcode == MMC_SEND_EXT_CSD && mmc_cmd_type(cmd) == MMC_CMD_ADTC))
-		rawcmd |= (0x1 << 11);
+		rawcmd |= BIT(11);
 
 	if (cmd->data) {
 		struct mmc_data *data = cmd->data;
@@ -1029,16 +1021,16 @@ static inline u32 msdc_cmd_prepare_raw_cmd(struct msdc_host *host,
 		if (mmc_op_multi(opcode)) {
 			if (mmc_card_mmc(mmc->card) && mrq->sbc &&
 			    !(mrq->sbc->arg & 0xFFFF0000))
-				rawcmd |= 0x2 << 28; /* AutoCMD23 */
+				rawcmd |= BIT(29); /* AutoCMD23 */
 		}
 
 		rawcmd |= ((data->blksz & 0xFFF) << 16);
 		if (data->flags & MMC_DATA_WRITE)
-			rawcmd |= (0x1 << 13);
+			rawcmd |= BIT(13);
 		if (data->blocks > 1)
-			rawcmd |= (0x2 << 11);
+			rawcmd |= BIT(12);
 		else
-			rawcmd |= (0x1 << 11);
+			rawcmd |= BIT(11);
 		/* Always use dma mode */
 		sdr_clr_bits(host->base + MSDC_CFG, MSDC_CFG_PIO);
 
@@ -1231,13 +1223,13 @@ static bool msdc_cmd_done(struct msdc_host *host, int events,
 static inline bool msdc_cmd_is_ready(struct msdc_host *host,
 		struct mmc_request *mrq, struct mmc_command *cmd)
 {
-	/* The max busy time we can endure is 20ms */
-	unsigned long tmo = jiffies + msecs_to_jiffies(20);
+	u32 val;
+	int ret;
 
-	while ((readl(host->base + SDC_STS) & SDC_STS_CMDBUSY) &&
-			time_before(jiffies, tmo))
-		cpu_relax();
-	if (readl(host->base + SDC_STS) & SDC_STS_CMDBUSY) {
+	/* The max busy time we can endure is 20ms */
+	ret = readl_poll_timeout_atomic(host->base + SDC_STS, val,
+					!(val & SDC_STS_CMDBUSY), 1, 20000);
+	if (ret) {
 		dev_err(host->dev, "CMD bus busy detected\n");
 		host->error |= REQ_CMD_BUSY;
 		msdc_cmd_done(host, MSDC_INT_CMDTMO, mrq, cmd);
@@ -1245,12 +1237,10 @@ static inline bool msdc_cmd_is_ready(struct msdc_host *host,
 	}
 
 	if (mmc_resp_type(cmd) == MMC_RSP_R1B || cmd->data) {
-		tmo = jiffies + msecs_to_jiffies(20);
 		/* R1B or with data, should check SDCBUSY */
-		while ((readl(host->base + SDC_STS) & SDC_STS_SDCBUSY) &&
-				time_before(jiffies, tmo))
-			cpu_relax();
-		if (readl(host->base + SDC_STS) & SDC_STS_SDCBUSY) {
+		ret = readl_poll_timeout_atomic(host->base + SDC_STS, val,
+						!(val & SDC_STS_SDCBUSY), 1, 20000);
+		if (ret) {
 			dev_err(host->dev, "Controller busy detected\n");
 			host->error |= REQ_CMD_BUSY;
 			msdc_cmd_done(host, MSDC_INT_CMDTMO, mrq, cmd);
@@ -1376,6 +1366,8 @@ static bool msdc_data_xfer_done(struct msdc_host *host, u32 events,
 	    (MSDC_INT_XFER_COMPL | MSDC_INT_DATCRCERR | MSDC_INT_DATTMO
 	     | MSDC_INT_DMA_BDCSERR | MSDC_INT_DMA_GPDCSERR
 	     | MSDC_INT_DMA_PROTECT);
+	u32 val;
+	int ret;
 
 	spin_lock_irqsave(&host->lock, flags);
 	done = !host->data;
@@ -1392,8 +1384,14 @@ static bool msdc_data_xfer_done(struct msdc_host *host, u32 events,
 				readl(host->base + MSDC_DMA_CFG));
 		sdr_set_field(host->base + MSDC_DMA_CTRL, MSDC_DMA_CTRL_STOP,
 				1);
-		while (readl(host->base + MSDC_DMA_CFG) & MSDC_DMA_CFG_STS)
-			cpu_relax();
+
+		ret = readl_poll_timeout_atomic(host->base + MSDC_DMA_CFG, val,
+						!(val & MSDC_DMA_CFG_STS), 1, 20000);
+		if (ret) {
+			dev_dbg(host->dev, "DMA stop timed out\n");
+			return false;
+		}
+
 		sdr_clr_bits(host->base + MSDC_INTEN, data_ints_mask);
 		dev_dbg(host->dev, "DMA stop\n");
 
@@ -1631,6 +1629,7 @@ static void msdc_init_hw(struct msdc_host *host)
 {
 	u32 val;
 	u32 tune_reg = host->dev_comp->pad_tune_reg;
+	struct mmc_host *mmc = mmc_from_priv(host);
 
 	if (host->reset) {
 		reset_control_assert(host->reset);
@@ -1685,7 +1684,7 @@ static void msdc_init_hw(struct msdc_host *host)
 	}
 
 	if (host->dev_comp->busy_check)
-		sdr_clr_bits(host->base + MSDC_PATCH_BIT1, (1 << 7));
+		sdr_clr_bits(host->base + MSDC_PATCH_BIT1, BIT(7));
 
 	if (host->dev_comp->async_fifo) {
 		sdr_set_field(host->base + MSDC_PATCH_BIT2,
@@ -1736,14 +1735,18 @@ static void msdc_init_hw(struct msdc_host *host)
 				     MSDC_PAD_TUNE_RXDLYSEL);
 	}
 
-	/* Configure to enable SDIO mode.
-	 * it's must otherwise sdio cmd5 failed
-	 */
-	sdr_set_bits(host->base + SDC_CFG, SDC_CFG_SDIO);
+	if (mmc->caps2 & MMC_CAP2_NO_SDIO) {
+		sdr_clr_bits(host->base + SDC_CFG, SDC_CFG_SDIO);
+		sdr_clr_bits(host->base + MSDC_INTEN, MSDC_INTEN_SDIOIRQ);
+		sdr_clr_bits(host->base + SDC_ADV_CFG0, SDC_DAT1_IRQ_TRIGGER);
+	} else {
+		/* Configure to enable SDIO mode, otherwise SDIO CMD5 fails */
+		sdr_set_bits(host->base + SDC_CFG, SDC_CFG_SDIO);
 
-	/* Config SDIO device detect interrupt function */
-	sdr_clr_bits(host->base + SDC_CFG, SDC_CFG_SDIOIDE);
-	sdr_set_bits(host->base + SDC_ADV_CFG0, SDC_DAT1_IRQ_TRIGGER);
+		/* Config SDIO device detect interrupt function */
+		sdr_clr_bits(host->base + SDC_CFG, SDC_CFG_SDIOIDE);
+		sdr_set_bits(host->base + SDC_ADV_CFG0, SDC_DAT1_IRQ_TRIGGER);
+	}
 
 	/* Configure to default data timeout */
 	sdr_set_field(host->base + SDC_CFG, SDC_CFG_DTOC, 3);
@@ -1865,7 +1868,7 @@ static void msdc_ops_set_ios(struct mmc_host *mmc, struct mmc_ios *ios)
 static u32 test_delay_bit(u32 delay, u32 bit)
 {
 	bit %= PAD_DELAY_MAX;
-	return delay & (1 << bit);
+	return delay & BIT(bit);
 }
 
 static int get_delay_len(u32 delay, u32 start_bit)
@@ -1970,9 +1973,9 @@ static int msdc_tune_response(struct mmc_host *mmc, u32 opcode)
 		for (j = 0; j < 3; j++) {
 			mmc_send_tuning(mmc, opcode, &cmd_err);
 			if (!cmd_err) {
-				rise_delay |= (1 << i);
+				rise_delay |= BIT(i);
 			} else {
-				rise_delay &= ~(1 << i);
+				rise_delay &= ~BIT(i);
 				break;
 			}
 		}
@@ -1994,9 +1997,9 @@ static int msdc_tune_response(struct mmc_host *mmc, u32 opcode)
 		for (j = 0; j < 3; j++) {
 			mmc_send_tuning(mmc, opcode, &cmd_err);
 			if (!cmd_err) {
-				fall_delay |= (1 << i);
+				fall_delay |= BIT(i);
 			} else {
-				fall_delay &= ~(1 << i);
+				fall_delay &= ~BIT(i);
 				break;
 			}
 		}
@@ -2024,7 +2027,7 @@ skip_fall:
 			      MSDC_PAD_TUNE_CMDRRDLY, i);
 		mmc_send_tuning(mmc, opcode, &cmd_err);
 		if (!cmd_err)
-			internal_delay |= (1 << i);
+			internal_delay |= BIT(i);
 	}
 	dev_dbg(host->dev, "Final internal delay: 0x%x\n", internal_delay);
 	internal_delay_phase = get_best_delay(host, internal_delay);
@@ -2069,9 +2072,9 @@ static int hs400_tune_response(struct mmc_host *mmc, u32 opcode)
 		for (j = 0; j < 3; j++) {
 			mmc_send_tuning(mmc, opcode, &cmd_err);
 			if (!cmd_err) {
-				cmd_delay |= (1 << i);
+				cmd_delay |= BIT(i);
 			} else {
-				cmd_delay &= ~(1 << i);
+				cmd_delay &= ~BIT(i);
 				break;
 			}
 		}
@@ -2101,7 +2104,7 @@ static int msdc_tune_data(struct mmc_host *mmc, u32 opcode)
 		msdc_set_data_delay(host, i);
 		ret = mmc_send_tuning(mmc, opcode, NULL);
 		if (!ret)
-			rise_delay |= (1 << i);
+			rise_delay |= BIT(i);
 	}
 	final_rise_delay = get_best_delay(host, rise_delay);
 	/* if rising edge has enough margin, then do not scan falling edge */
@@ -2115,7 +2118,7 @@ static int msdc_tune_data(struct mmc_host *mmc, u32 opcode)
 		msdc_set_data_delay(host, i);
 		ret = mmc_send_tuning(mmc, opcode, NULL);
 		if (!ret)
-			fall_delay |= (1 << i);
+			fall_delay |= BIT(i);
 	}
 	final_fall_delay = get_best_delay(host, fall_delay);
 
@@ -2159,7 +2162,7 @@ static int msdc_tune_together(struct mmc_host *mmc, u32 opcode)
 		msdc_set_data_delay(host, i);
 		ret = mmc_send_tuning(mmc, opcode, NULL);
 		if (!ret)
-			rise_delay |= (1 << i);
+			rise_delay |= BIT(i);
 	}
 	final_rise_delay = get_best_delay(host, rise_delay);
 	/* if rising edge has enough margin, then do not scan falling edge */
@@ -2175,7 +2178,7 @@ static int msdc_tune_together(struct mmc_host *mmc, u32 opcode)
 		msdc_set_data_delay(host, i);
 		ret = mmc_send_tuning(mmc, opcode, NULL);
 		if (!ret)
-			fall_delay |= (1 << i);
+			fall_delay |= BIT(i);
 	}
 	final_fall_delay = get_best_delay(host, fall_delay);
 
@@ -2292,7 +2295,7 @@ static int msdc_execute_hs400_tuning(struct mmc_host *mmc, struct mmc_card *card
 				      PAD_DS_TUNE_DLY1, i);
 		ret = mmc_get_ext_csd(card, &ext_csd);
 		if (!ret) {
-			result_dly1 |= (1 << i);
+			result_dly1 |= BIT(i);
 			kfree(ext_csd);
 		}
 	}
@@ -2516,7 +2519,20 @@ static int msdc_of_clock_parse(struct platform_device *pdev,
 	/*source clock control gate is optional clock*/
 	host->src_clk_cg = devm_clk_get_optional(&pdev->dev, "source_cg");
 	if (IS_ERR(host->src_clk_cg))
-		host->src_clk_cg = NULL;
+		return PTR_ERR(host->src_clk_cg);
+
+	/*
+	 * Fallback for legacy device-trees: src_clk and HCLK use the same
+	 * bit to control gating but they are parented to a different mux,
+	 * hence if our intention is to gate only the source, required
+	 * during a clk mode switch to avoid hw hangs, we need to gate
+	 * its parent (specified as a different clock only on new DTs).
+	 */
+	if (!host->src_clk_cg) {
+		host->src_clk_cg = clk_get_parent(host->src_clk);
+		if (IS_ERR(host->src_clk_cg))
+			return PTR_ERR(host->src_clk_cg);
+	}
 
 	host->sys_clk_cg = devm_clk_get_optional(&pdev->dev, "sys_cg");
 	if (IS_ERR(host->sys_clk_cg))
@@ -2674,7 +2690,11 @@ static int msdc_drv_probe(struct platform_device *pdev)
 	spin_lock_init(&host->lock);
 
 	platform_set_drvdata(pdev, mmc);
-	msdc_ungate_clock(host);
+	ret = msdc_ungate_clock(host);
+	if (ret) {
+		dev_err(&pdev->dev, "Cannot ungate clocks!\n");
+		goto release_mem;
+	}
 	msdc_init_hw(host);
 
 	if (mmc->caps2 & MMC_CAP2_CQE) {
@@ -2833,8 +2853,12 @@ static int __maybe_unused msdc_runtime_resume(struct device *dev)
 {
 	struct mmc_host *mmc = dev_get_drvdata(dev);
 	struct msdc_host *host = mmc_priv(mmc);
+	int ret;
+
+	ret = msdc_ungate_clock(host);
+	if (ret)
+		return ret;
 
-	msdc_ungate_clock(host);
 	msdc_restore_reg(host);
 	return 0;
 }
diff --git a/drivers/mmc/host/omap_hsmmc.c b/drivers/mmc/host/omap_hsmmc.c
index 9dafcbf969d9..fca30add563e 100644
--- a/drivers/mmc/host/omap_hsmmc.c
+++ b/drivers/mmc/host/omap_hsmmc.c
@@ -1499,41 +1499,6 @@ static void omap_hsmmc_set_ios(struct mmc_host *mmc, struct mmc_ios *ios)
 	omap_hsmmc_set_bus_mode(host);
 }
 
-static void omap_hsmmc_init_card(struct mmc_host *mmc, struct mmc_card *card)
-{
-	struct omap_hsmmc_host *host = mmc_priv(mmc);
-
-	if (card->type == MMC_TYPE_SDIO || card->type == MMC_TYPE_SD_COMBO) {
-		struct device_node *np = mmc_dev(mmc)->of_node;
-
-		/*
-		 * REVISIT: should be moved to sdio core and made more
-		 * general e.g. by expanding the DT bindings of child nodes
-		 * to provide a mechanism to provide this information:
-		 * Documentation/devicetree/bindings/mmc/mmc-card.yaml
-		 */
-
-		np = of_get_compatible_child(np, "ti,wl1251");
-		if (np) {
-			/*
-			 * We have TI wl1251 attached to MMC3. Pass this
-			 * information to the SDIO core because it can't be
-			 * probed by normal methods.
-			 */
-
-			dev_info(host->dev, "found wl1251\n");
-			card->quirks |= MMC_QUIRK_NONSTD_SDIO;
-			card->cccr.wide_bus = 1;
-			card->cis.vendor = 0x104c;
-			card->cis.device = 0x9066;
-			card->cis.blksize = 512;
-			card->cis.max_dtr = 24000000;
-			card->ocr = 0x80;
-			of_node_put(np);
-		}
-	}
-}
-
 static void omap_hsmmc_enable_sdio_irq(struct mmc_host *mmc, int enable)
 {
 	struct omap_hsmmc_host *host = mmc_priv(mmc);
@@ -1660,7 +1625,6 @@ static struct mmc_host_ops omap_hsmmc_ops = {
 	.set_ios = omap_hsmmc_set_ios,
 	.get_cd = mmc_gpio_get_cd,
 	.get_ro = mmc_gpio_get_ro,
-	.init_card = omap_hsmmc_init_card,
 	.enable_sdio_irq = omap_hsmmc_enable_sdio_irq,
 };
 
diff --git a/drivers/mmc/host/sdhci-esdhc-imx.c b/drivers/mmc/host/sdhci-esdhc-imx.c
index 764ee1b761d9..55981b0f0b10 100644
--- a/drivers/mmc/host/sdhci-esdhc-imx.c
+++ b/drivers/mmc/host/sdhci-esdhc-imx.c
@@ -305,6 +305,9 @@ static struct esdhc_soc_data usdhc_imx7ulp_data = {
 			| ESDHC_FLAG_PMQOS | ESDHC_FLAG_HS400
 			| ESDHC_FLAG_STATE_LOST_IN_LPMODE,
 };
+static struct esdhc_soc_data usdhc_imxrt1050_data = {
+	.flags = ESDHC_FLAG_USDHC | ESDHC_FLAG_HS200 | ESDHC_FLAG_ERR004536,
+};
 
 static struct esdhc_soc_data usdhc_imx8qxp_data = {
 	.flags = ESDHC_FLAG_USDHC | ESDHC_FLAG_STD_TUNING
@@ -355,6 +358,7 @@ static const struct of_device_id imx_esdhc_dt_ids[] = {
 	{ .compatible = "fsl,imx7ulp-usdhc", .data = &usdhc_imx7ulp_data, },
 	{ .compatible = "fsl,imx8qxp-usdhc", .data = &usdhc_imx8qxp_data, },
 	{ .compatible = "fsl,imx8mm-usdhc", .data = &usdhc_imx8mm_data, },
+	{ .compatible = "fsl,imxrt1050-usdhc", .data = &usdhc_imxrt1050_data, },
 	{ .compatible = "nxp,s32g2-usdhc", .data = &usdhc_s32g2_data, },
 	{ /* sentinel */ }
 };
diff --git a/drivers/mmc/host/sdhci-pci-core.c b/drivers/mmc/host/sdhci-pci-core.c
index 6f9877546830..ed53276f6ad9 100644
--- a/drivers/mmc/host/sdhci-pci-core.c
+++ b/drivers/mmc/host/sdhci-pci-core.c
@@ -1866,6 +1866,7 @@ static const struct pci_device_id pci_ids[] = {
 	SDHCI_PCI_DEVICE(INTEL, JSL_SD,    intel_byt_sd),
 	SDHCI_PCI_DEVICE(INTEL, LKF_EMMC,  intel_glk_emmc),
 	SDHCI_PCI_DEVICE(INTEL, LKF_SD,    intel_byt_sd),
+	SDHCI_PCI_DEVICE(INTEL, ADL_EMMC,  intel_glk_emmc),
 	SDHCI_PCI_DEVICE(O2, 8120,     o2),
 	SDHCI_PCI_DEVICE(O2, 8220,     o2),
 	SDHCI_PCI_DEVICE(O2, 8221,     o2),
diff --git a/drivers/mmc/host/sdhci-pci-gli.c b/drivers/mmc/host/sdhci-pci-gli.c
index 4fd99c1e82ba..97035d77c18c 100644
--- a/drivers/mmc/host/sdhci-pci-gli.c
+++ b/drivers/mmc/host/sdhci-pci-gli.c
@@ -12,6 +12,7 @@
 #include <linux/pci.h>
 #include <linux/mmc/mmc.h>
 #include <linux/delay.h>
+#include <linux/of.h>
 #include "sdhci.h"
 #include "sdhci-pci.h"
 #include "cqhci.h"
@@ -116,6 +117,8 @@
 #define PCI_GLI_9755_PECONF   0x44
 #define   PCI_GLI_9755_LFCLK    GENMASK(14, 12)
 #define   PCI_GLI_9755_DMACLK   BIT(29)
+#define   PCI_GLI_9755_INVERT_CD  BIT(30)
+#define   PCI_GLI_9755_INVERT_WP  BIT(31)
 
 #define PCI_GLI_9755_CFG2          0x48
 #define   PCI_GLI_9755_CFG2_L1DLY    GENMASK(28, 24)
@@ -570,6 +573,14 @@ static void gl9755_hw_setting(struct sdhci_pci_slot *slot)
 	gl9755_wt_on(pdev);
 
 	pci_read_config_dword(pdev, PCI_GLI_9755_PECONF, &value);
+	/*
+	 * Apple ARM64 platforms using these chips may have
+	 * inverted CD/WP detection.
+	 */
+	if (of_property_read_bool(pdev->dev.of_node, "cd-inverted"))
+		value |= PCI_GLI_9755_INVERT_CD;
+	if (of_property_read_bool(pdev->dev.of_node, "wp-inverted"))
+		value |= PCI_GLI_9755_INVERT_WP;
 	value &= ~PCI_GLI_9755_LFCLK;
 	value &= ~PCI_GLI_9755_DMACLK;
 	pci_write_config_dword(pdev, PCI_GLI_9755_PECONF, value);
@@ -891,7 +902,28 @@ static int gli_probe_slot_gl9763e(struct sdhci_pci_slot *slot)
 	return 0;
 }
 
+#define REG_OFFSET_IN_BITS(reg) ((reg) << 3 & 0x18)
+
+static u16 sdhci_gli_readw(struct sdhci_host *host, int reg)
+{
+	u32 val = readl(host->ioaddr + (reg & ~3));
+	u16 word;
+
+	word = (val >> REG_OFFSET_IN_BITS(reg)) & 0xffff;
+	return word;
+}
+
+static u8 sdhci_gli_readb(struct sdhci_host *host, int reg)
+{
+	u32 val = readl(host->ioaddr + (reg & ~3));
+	u8 byte = (val >> REG_OFFSET_IN_BITS(reg)) & 0xff;
+
+	return byte;
+}
+
 static const struct sdhci_ops sdhci_gl9755_ops = {
+	.read_w			= sdhci_gli_readw,
+	.read_b			= sdhci_gli_readb,
 	.set_clock		= sdhci_gl9755_set_clock,
 	.enable_dma		= sdhci_pci_enable_dma,
 	.set_bus_width		= sdhci_set_bus_width,
@@ -911,6 +943,8 @@ const struct sdhci_pci_fixes sdhci_gl9755 = {
 };
 
 static const struct sdhci_ops sdhci_gl9750_ops = {
+	.read_w			= sdhci_gli_readw,
+	.read_b			= sdhci_gli_readb,
 	.read_l                 = sdhci_gl9750_readl,
 	.set_clock		= sdhci_gl9750_set_clock,
 	.enable_dma		= sdhci_pci_enable_dma,
diff --git a/drivers/mmc/host/sdhci-pci-o2micro.c b/drivers/mmc/host/sdhci-pci-o2micro.c
index f045c1ee4667..92c20cb8074a 100644
--- a/drivers/mmc/host/sdhci-pci-o2micro.c
+++ b/drivers/mmc/host/sdhci-pci-o2micro.c
@@ -12,6 +12,7 @@
 #include <linux/mmc/mmc.h>
 #include <linux/delay.h>
 #include <linux/iopoll.h>
+#include <linux/bitfield.h>
 
 #include "sdhci.h"
 #include "sdhci-pci.h"
@@ -43,12 +44,16 @@
 #define O2_SD_CAP_REG0		0x334
 #define O2_SD_UHS1_CAP_SETTING	0x33C
 #define O2_SD_DELAY_CTRL	0x350
+#define O2_SD_OUTPUT_CLK_SOURCE_SWITCH	0x354
 #define O2_SD_UHS2_L1_CTRL	0x35C
 #define O2_SD_FUNC_REG3		0x3E0
 #define O2_SD_FUNC_REG4		0x3E4
 #define O2_SD_LED_ENABLE	BIT(6)
 #define O2_SD_FREG0_LEDOFF	BIT(13)
+#define O2_SD_SEL_DLL		BIT(16)
 #define O2_SD_FREG4_ENABLE_CLK_SET	BIT(22)
+#define O2_SD_PHASE_MASK	GENMASK(23, 20)
+#define O2_SD_FIX_PHASE		FIELD_PREP(O2_SD_PHASE_MASK, 0x9)
 
 #define O2_SD_VENDOR_SETTING	0x110
 #define O2_SD_VENDOR_SETTING2	0x1C8
@@ -301,9 +306,13 @@ static int sdhci_o2_dll_recovery(struct sdhci_host *host)
 static int sdhci_o2_execute_tuning(struct mmc_host *mmc, u32 opcode)
 {
 	struct sdhci_host *host = mmc_priv(mmc);
+	struct sdhci_pci_slot *slot = sdhci_priv(host);
+	struct sdhci_pci_chip *chip = slot->chip;
 	int current_bus_width = 0;
 	u32 scratch32 = 0;
 	u16 scratch = 0;
+	u8  scratch_8 = 0;
+	u32 reg_val;
 
 	/*
 	 * This handler only implements the eMMC tuning that is specific to
@@ -322,6 +331,32 @@ static int sdhci_o2_execute_tuning(struct mmc_host *mmc, u32 opcode)
 	scratch |= O2_SD_PWR_FORCE_L0;
 	sdhci_writew(host, scratch, O2_SD_MISC_CTRL);
 
+	/* Stop clk */
+	reg_val = sdhci_readw(host, SDHCI_CLOCK_CONTROL);
+	reg_val &= ~SDHCI_CLOCK_CARD_EN;
+	sdhci_writew(host, reg_val, SDHCI_CLOCK_CONTROL);
+
+	/* UnLock WP */
+	pci_read_config_byte(chip->pdev, O2_SD_LOCK_WP, &scratch_8);
+	scratch_8 &= 0x7f;
+	pci_write_config_byte(chip->pdev, O2_SD_LOCK_WP, scratch_8);
+
+	/* Set pcr 0x354[16] to choose dll clock, and set the default phase */
+	pci_read_config_dword(chip->pdev, O2_SD_OUTPUT_CLK_SOURCE_SWITCH, &reg_val);
+	reg_val &= ~(O2_SD_SEL_DLL | O2_SD_PHASE_MASK);
+	reg_val |= (O2_SD_SEL_DLL | O2_SD_FIX_PHASE);
+	pci_write_config_dword(chip->pdev, O2_SD_OUTPUT_CLK_SOURCE_SWITCH, reg_val);
+
+	/* Lock WP */
+	pci_read_config_byte(chip->pdev, O2_SD_LOCK_WP, &scratch_8);
+	scratch_8 |= 0x80;
+	pci_write_config_byte(chip->pdev, O2_SD_LOCK_WP, scratch_8);
+
+	/* Start clk */
+	reg_val = sdhci_readw(host, SDHCI_CLOCK_CONTROL);
+	reg_val |= SDHCI_CLOCK_CARD_EN;
+	sdhci_writew(host, reg_val, SDHCI_CLOCK_CONTROL);
+
 	/* wait DLL lock, timeout value 5ms */
 	if (readx_poll_timeout(sdhci_o2_pll_dll_wdt_control, host,
 		scratch32, (scratch32 & O2_DLL_LOCK_STATUS), 1, 5000))
@@ -533,23 +568,32 @@ static void sdhci_pci_o2_set_clock(struct sdhci_host *host, unsigned int clock)
 	if (clock == 0)
 		return;
 
-	if ((host->timing == MMC_TIMING_UHS_SDR104) && (clock == 200000000)) {
-		pci_read_config_byte(chip->pdev, O2_SD_LOCK_WP, &scratch);
-
-		scratch &= 0x7f;
-		pci_write_config_byte(chip->pdev, O2_SD_LOCK_WP, scratch);
+	/* UnLock WP */
+	pci_read_config_byte(chip->pdev, O2_SD_LOCK_WP, &scratch);
+	scratch &= 0x7f;
+	pci_write_config_byte(chip->pdev, O2_SD_LOCK_WP, scratch);
 
+	if ((host->timing == MMC_TIMING_UHS_SDR104) && (clock == 200000000)) {
 		pci_read_config_dword(chip->pdev, O2_SD_PLL_SETTING, &scratch_32);
 
 		if ((scratch_32 & 0xFFFF0000) != 0x2c280000)
 			o2_pci_set_baseclk(chip, 0x2c280000);
+	} else {
+		pci_read_config_dword(chip->pdev, O2_SD_PLL_SETTING, &scratch_32);
 
-		pci_read_config_byte(chip->pdev, O2_SD_LOCK_WP, &scratch);
-
-		scratch |= 0x80;
-		pci_write_config_byte(chip->pdev, O2_SD_LOCK_WP, scratch);
+		if ((scratch_32 & 0xFFFF0000) != 0x25100000)
+			o2_pci_set_baseclk(chip, 0x25100000);
 	}
 
+	pci_read_config_dword(chip->pdev, O2_SD_OUTPUT_CLK_SOURCE_SWITCH, &scratch_32);
+	scratch_32 &= ~(O2_SD_SEL_DLL | O2_SD_PHASE_MASK);
+	pci_write_config_dword(chip->pdev, O2_SD_OUTPUT_CLK_SOURCE_SWITCH, scratch_32);
+
+	/* Lock WP */
+	pci_read_config_byte(chip->pdev, O2_SD_LOCK_WP, &scratch);
+	scratch |= 0x80;
+	pci_write_config_byte(chip->pdev, O2_SD_LOCK_WP, scratch);
+
 	clk = sdhci_calc_clk(host, clock, &host->mmc->actual_clock);
 	sdhci_o2_enable_clk(host, clk);
 }
diff --git a/drivers/mmc/host/sdhci-pci.h b/drivers/mmc/host/sdhci-pci.h
index 5e3193278ff9..3661a224fb04 100644
--- a/drivers/mmc/host/sdhci-pci.h
+++ b/drivers/mmc/host/sdhci-pci.h
@@ -59,6 +59,7 @@
 #define PCI_DEVICE_ID_INTEL_JSL_SD	0x4df8
 #define PCI_DEVICE_ID_INTEL_LKF_EMMC	0x98c4
 #define PCI_DEVICE_ID_INTEL_LKF_SD	0x98f8
+#define PCI_DEVICE_ID_INTEL_ADL_EMMC	0x54c4
 
 #define PCI_DEVICE_ID_SYSKONNECT_8000	0x8000
 #define PCI_DEVICE_ID_VIA_95D0		0x95d0
diff --git a/drivers/mmc/host/tmio_mmc_core.c b/drivers/mmc/host/tmio_mmc_core.c
index e2affa52ef46..a5850d83908b 100644
--- a/drivers/mmc/host/tmio_mmc_core.c
+++ b/drivers/mmc/host/tmio_mmc_core.c
@@ -960,14 +960,8 @@ static void tmio_mmc_set_ios(struct mmc_host *mmc, struct mmc_ios *ios)
 	case MMC_POWER_OFF:
 		tmio_mmc_power_off(host);
 		/* For R-Car Gen2+, we need to reset SDHI specific SCC */
-		if (host->pdata->flags & TMIO_MMC_MIN_RCAR2) {
-			host->reset(host);
-
-			if (host->native_hotplug)
-				tmio_mmc_enable_mmc_irqs(host,
-						TMIO_STAT_CARD_REMOVE |
-						TMIO_STAT_CARD_INSERT);
-		}
+		if (host->pdata->flags & TMIO_MMC_MIN_RCAR2)
+			tmio_mmc_reset(host);
 
 		host->set_clock(host, 0);
 		break;
@@ -1175,6 +1169,7 @@ int tmio_mmc_host_probe(struct tmio_mmc_host *_host)
 	if (mmc_can_gpio_cd(mmc))
 		_host->ops.get_cd = mmc_gpio_get_cd;
 
+	/* must be set before tmio_mmc_reset() */
 	_host->native_hotplug = !(mmc_can_gpio_cd(mmc) ||
 				  mmc->caps & MMC_CAP_NEEDS_POLL ||
 				  !mmc_card_is_removable(mmc));
@@ -1295,10 +1290,6 @@ int tmio_mmc_host_runtime_resume(struct device *dev)
 	if (host->clk_cache)
 		host->set_clock(host, host->clk_cache);
 
-	if (host->native_hotplug)
-		tmio_mmc_enable_mmc_irqs(host,
-				TMIO_STAT_CARD_REMOVE | TMIO_STAT_CARD_INSERT);
-
 	tmio_mmc_enable_dma(host, true);
 
 	return 0;