Merge tag 'mmc-2-20-2020' of https://gitlab.denx.de/u-boot/custodians/u-boot-mmc

- sdhci: code clean-up and fix cache coherency problem
- enable cache snooping on mpc830x
- Fix build error when MMC_WRITE disabled

32 files changed, 187 insertions, 192 deletions

diff --git a/arch/arc/include/asm/dma-mapping.h b/arch/arc/include/asm/dma-mapping.h
new file mode 100644
index 0000000000..853b0877b3
--- /dev/null
+++ b/arch/arc/include/asm/dma-mapping.h
@@ -0,0 +1 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
diff --git a/arch/arm/include/asm/dma-mapping.h b/arch/arm/include/asm/dma-mapping.h
index d20703739f..8bb3fa2e0e 100644
--- a/arch/arm/include/asm/dma-mapping.h
+++ b/arch/arm/include/asm/dma-mapping.h
@@ -11,10 +11,9 @@
 #include <asm/cache.h>
 #include <cpu_func.h>
 #include <linux/dma-direction.h>
+#include <linux/types.h>
 #include <malloc.h>
 
-#define	dma_mapping_error(x, y)	0
-
 static inline void *dma_alloc_coherent(size_t len, unsigned long *handle)
 {
 	*handle = (unsigned long)memalign(ARCH_DMA_MINALIGN, ROUND(len, ARCH_DMA_MINALIGN));
@@ -26,30 +25,4 @@ static inline void dma_free_coherent(void *addr)
 	free(addr);
 }
 
-static inline unsigned long dma_map_single(volatile void *vaddr, size_t len,
-					   enum dma_data_direction dir)
-{
-	unsigned long addr = (unsigned long)vaddr;
-
-	len = ALIGN(len, ARCH_DMA_MINALIGN);
-
-	if (dir == DMA_FROM_DEVICE)
-		invalidate_dcache_range(addr, addr + len);
-	else
-		flush_dcache_range(addr, addr + len);
-
-	return addr;
-}
-
-static inline void dma_unmap_single(volatile void *vaddr, size_t len,
-				    enum dma_data_direction dir)
-{
-	unsigned long addr = (unsigned long)vaddr;
-
-	len = ALIGN(len, ARCH_DMA_MINALIGN);
-
-	if (dir != DMA_TO_DEVICE)
-		invalidate_dcache_range(addr, addr + len);
-}
-
 #endif /* __ASM_ARM_DMA_MAPPING_H */
diff --git a/arch/m68k/include/asm/dma-mapping.h b/arch/m68k/include/asm/dma-mapping.h
new file mode 100644
index 0000000000..853b0877b3
--- /dev/null
+++ b/arch/m68k/include/asm/dma-mapping.h
@@ -0,0 +1 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
diff --git a/arch/microblaze/include/asm/dma-mapping.h b/arch/microblaze/include/asm/dma-mapping.h
new file mode 100644
index 0000000000..853b0877b3
--- /dev/null
+++ b/arch/microblaze/include/asm/dma-mapping.h
@@ -0,0 +1 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
diff --git a/arch/mips/include/asm/dma-mapping.h b/arch/mips/include/asm/dma-mapping.h
new file mode 100644
index 0000000000..853b0877b3
--- /dev/null
+++ b/arch/mips/include/asm/dma-mapping.h
@@ -0,0 +1 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
diff --git a/arch/nds32/include/asm/dma-mapping.h b/arch/nds32/include/asm/dma-mapping.h
index c8876ceadd..f8668f1fe5 100644
--- a/arch/nds32/include/asm/dma-mapping.h
+++ b/arch/nds32/include/asm/dma-mapping.h
@@ -10,6 +10,7 @@
 #include <asm/cache.h>
 #include <cpu_func.h>
 #include <linux/dma-direction.h>
+#include <linux/types.h>
 #include <malloc.h>
 
 static void *dma_alloc_coherent(size_t len, unsigned long *handle)
@@ -18,30 +19,4 @@ static void *dma_alloc_coherent(size_t len, unsigned long *handle)
 	return (void *)*handle;
 }
 
-static inline unsigned long dma_map_single(volatile void *vaddr, size_t len,
-					   enum dma_data_direction dir)
-{
-	unsigned long addr = (unsigned long)vaddr;
-
-	len = ALIGN(len, ARCH_DMA_MINALIGN);
-
-	if (dir == DMA_FROM_DEVICE)
-		invalidate_dcache_range(addr, addr + len);
-	else
-		flush_dcache_range(addr, addr + len);
-
-	return addr;
-}
-
-static inline void dma_unmap_single(volatile void *vaddr, size_t len,
-				    enum dma_data_direction dir)
-{
-	unsigned long addr = (unsigned long)vaddr;
-
-	len = ALIGN(len, ARCH_DMA_MINALIGN);
-
-	if (dir != DMA_TO_DEVICE)
-		invalidate_dcache_range(addr, addr + len);
-}
-
 #endif /* __ASM_NDS_DMA_MAPPING_H */
diff --git a/arch/powerpc/include/asm/dma-mapping.h b/arch/powerpc/include/asm/dma-mapping.h
new file mode 100644
index 0000000000..853b0877b3
--- /dev/null
+++ b/arch/powerpc/include/asm/dma-mapping.h
@@ -0,0 +1 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
diff --git a/arch/riscv/include/asm/dma-mapping.h b/arch/riscv/include/asm/dma-mapping.h
index 6cc3946959..6ecadab41c 100644
--- a/arch/riscv/include/asm/dma-mapping.h
+++ b/arch/riscv/include/asm/dma-mapping.h
@@ -10,13 +10,12 @@
 #define __ASM_RISCV_DMA_MAPPING_H
 
 #include <common.h>
+#include <linux/types.h>
 #include <asm/cache.h>
 #include <cpu_func.h>
 #include <linux/dma-direction.h>
 #include <malloc.h>
 
-#define dma_mapping_error(x, y)	0
-
 static inline void *dma_alloc_coherent(size_t len, unsigned long *handle)
 {
 	*handle = (unsigned long)memalign(ARCH_DMA_MINALIGN, len);
@@ -28,30 +27,4 @@ static inline void dma_free_coherent(void *addr)
 	free(addr);
 }
 
-static inline unsigned long dma_map_single(volatile void *vaddr, size_t len,
-					   enum dma_data_direction dir)
-{
-	unsigned long addr = (unsigned long)vaddr;
-
-	len = ALIGN(len, ARCH_DMA_MINALIGN);
-
-	if (dir == DMA_FROM_DEVICE)
-		invalidate_dcache_range(addr, addr + len);
-	else
-		flush_dcache_range(addr, addr + len);
-
-	return addr;
-}
-
-static inline void dma_unmap_single(volatile void *vaddr, size_t len,
-				    enum dma_data_direction dir)
-{
-	unsigned long addr = (unsigned long)vaddr;
-
-	len = ALIGN(len, ARCH_DMA_MINALIGN);
-
-	if (dir != DMA_TO_DEVICE)
-		invalidate_dcache_range(addr, addr + len);
-}
-
 #endif /* __ASM_RISCV_DMA_MAPPING_H */
diff --git a/arch/sandbox/include/asm/dma-mapping.h b/arch/sandbox/include/asm/dma-mapping.h
new file mode 100644
index 0000000000..853b0877b3
--- /dev/null
+++ b/arch/sandbox/include/asm/dma-mapping.h
@@ -0,0 +1 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
diff --git a/arch/sh/include/asm/dma-mapping.h b/arch/sh/include/asm/dma-mapping.h
new file mode 100644
index 0000000000..853b0877b3
--- /dev/null
+++ b/arch/sh/include/asm/dma-mapping.h
@@ -0,0 +1 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
diff --git a/arch/x86/include/asm/dma-mapping.h b/arch/x86/include/asm/dma-mapping.h
index 900b99b8a6..8be1003e6b 100644
--- a/arch/x86/include/asm/dma-mapping.h
+++ b/arch/x86/include/asm/dma-mapping.h
@@ -11,10 +11,9 @@
 #include <asm/cache.h>
 #include <cpu_func.h>
 #include <linux/dma-direction.h>
+#include <linux/types.h>
 #include <malloc.h>
 
-#define	dma_mapping_error(x, y)	0
-
 static inline void *dma_alloc_coherent(size_t len, unsigned long *handle)
 {
 	*handle = (unsigned long)memalign(ARCH_DMA_MINALIGN, len);
@@ -26,30 +25,4 @@ static inline void dma_free_coherent(void *addr)
 	free(addr);
 }
 
-static inline unsigned long dma_map_single(volatile void *vaddr, size_t len,
-					   enum dma_data_direction dir)
-{
-	unsigned long addr = (unsigned long)vaddr;
-
-	len = ALIGN(len, ARCH_DMA_MINALIGN);
-
-	if (dir == DMA_FROM_DEVICE)
-		invalidate_dcache_range(addr, addr + len);
-	else
-		flush_dcache_range(addr, addr + len);
-
-	return addr;
-}
-
-static inline void dma_unmap_single(volatile void *vaddr, size_t len,
-				    enum dma_data_direction dir)
-{
-	unsigned long addr = (unsigned long)vaddr;
-
-	len = ALIGN(len, ARCH_DMA_MINALIGN);
-
-	if (dir != DMA_TO_DEVICE)
-		invalidate_dcache_range(addr, addr + len);
-}
-
 #endif /* __ASM_X86_DMA_MAPPING_H */
diff --git a/arch/xtensa/include/asm/dma-mapping.h b/arch/xtensa/include/asm/dma-mapping.h
new file mode 100644
index 0000000000..853b0877b3
--- /dev/null
+++ b/arch/xtensa/include/asm/dma-mapping.h
@@ -0,0 +1 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
diff --git a/drivers/dma/ti/k3-udma.c b/drivers/dma/ti/k3-udma.c
index f274100f32..e587f1fcb2 100644
--- a/drivers/dma/ti/k3-udma.c
+++ b/drivers/dma/ti/k3-udma.c
@@ -10,7 +10,7 @@
 #include <asm/io.h>
 #include <asm/bitops.h>
 #include <malloc.h>
-#include <asm/dma-mapping.h>
+#include <linux/dma-mapping.h>
 #include <dm.h>
 #include <dm/device_compat.h>
 #include <dm/devres.h>
diff --git a/drivers/mmc/fsl_esdhc.c b/drivers/mmc/fsl_esdhc.c
index ab40019ccb..386781da03 100644
--- a/drivers/mmc/fsl_esdhc.c
+++ b/drivers/mmc/fsl_esdhc.c
@@ -578,6 +578,18 @@ static int esdhc_set_ios_common(struct fsl_esdhc_priv *priv, struct mmc *mmc)
 	return 0;
 }
 
+static void esdhc_enable_cache_snooping(struct fsl_esdhc *regs)
+{
+#ifdef CONFIG_ARCH_MPC830X
+	immap_t *immr = (immap_t *)CONFIG_SYS_IMMR;
+	sysconf83xx_t *sysconf = &immr->sysconf;
+
+	setbits_be32(&sysconf->sdhccr, 0x02000000);
+#else
+	esdhc_write32(&regs->esdhcctl, 0x00000040);
+#endif
+}
+
 static int esdhc_init_common(struct fsl_esdhc_priv *priv, struct mmc *mmc)
 {
 	struct fsl_esdhc *regs = priv->esdhc_regs;
@@ -593,8 +605,7 @@ static int esdhc_init_common(struct fsl_esdhc_priv *priv, struct mmc *mmc)
 			return -ETIMEDOUT;
 	}
 
-	/* Enable cache snooping */
-	esdhc_write32(&regs->esdhcctl, 0x00000040);
+	esdhc_enable_cache_snooping(regs);
 
 	esdhc_setbits32(&regs->sysctl, SYSCTL_HCKEN | SYSCTL_IPGEN);
 
diff --git a/drivers/mmc/mmc.c b/drivers/mmc/mmc.c
index d43983d4a6..b50fcbf6cf 100644
--- a/drivers/mmc/mmc.c
+++ b/drivers/mmc/mmc.c
@@ -24,10 +24,6 @@
 #define DEFAULT_CMD6_TIMEOUT_MS  500
 
 static int mmc_set_signal_voltage(struct mmc *mmc, uint signal_voltage);
-static int mmc_power_cycle(struct mmc *mmc);
-#if !CONFIG_IS_ENABLED(MMC_TINY)
-static int mmc_select_mode_and_width(struct mmc *mmc, uint card_caps);
-#endif
 
 #if !CONFIG_IS_ENABLED(DM_MMC)
 
@@ -1126,9 +1122,11 @@ int mmc_hwpart_config(struct mmc *mmc,
 
 		ext_csd[EXT_CSD_ERASE_GROUP_DEF] = 1;
 
+#if CONFIG_IS_ENABLED(MMC_WRITE)
 		/* update erase group size to be high-capacity */
 		mmc->erase_grp_size =
 			ext_csd[EXT_CSD_HC_ERASE_GRP_SIZE] * 1024;
+#endif
 
 	}
 
@@ -2575,7 +2573,7 @@ static int mmc_startup(struct mmc *mmc)
 		err = mmc_get_capabilities(mmc);
 		if (err)
 			return err;
-		mmc_select_mode_and_width(mmc, mmc->card_caps);
+		err = mmc_select_mode_and_width(mmc, mmc->card_caps);
 	}
 #endif
 	if (err)
diff --git a/drivers/mmc/rpmb.c b/drivers/mmc/rpmb.c
index 33371fe562..ee6dbe30db 100644
--- a/drivers/mmc/rpmb.c
+++ b/drivers/mmc/rpmb.c
@@ -11,6 +11,7 @@
 #include <common.h>
 #include <memalign.h>
 #include <mmc.h>
+#include <sdhci.h>
 #include <u-boot/sha256.h>
 #include "mmc_private.h"
 
@@ -91,6 +92,7 @@ static int mmc_rpmb_request(struct mmc *mmc, const struct s_rpmb *s,
 {
 	struct mmc_cmd cmd = {0};
 	struct mmc_data data;
+	struct sdhci_host *host = mmc->priv;
 	int ret;
 
 	ret = mmc_set_blockcount(mmc, count, is_rel_write);
@@ -105,6 +107,9 @@ static int mmc_rpmb_request(struct mmc *mmc, const struct s_rpmb *s,
 	cmd.cmdarg = 0;
 	cmd.resp_type = MMC_RSP_R1;
 
+	if (host->quirks & SDHCI_QUIRK_BROKEN_R1B)
+		cmd.resp_type = MMC_RSP_R1;
+
 	data.src = (const char *)s;
 	data.blocks = 1;
 	data.blocksize = MMC_MAX_BLOCK_LEN;
diff --git a/drivers/mmc/sdhci-cadence.c b/drivers/mmc/sdhci-cadence.c
index e9108dabd1..2d38c2270e 100644
--- a/drivers/mmc/sdhci-cadence.c
+++ b/drivers/mmc/sdhci-cadence.c
@@ -171,6 +171,7 @@ static int sdhci_cdns_set_tune_val(struct sdhci_cdns_plat *plat,
 {
 	void __iomem *reg = plat->hrs_addr + SDHCI_CDNS_HRS06;
 	u32 tmp;
+	int i, ret;
 
 	if (WARN_ON(!FIELD_FIT(SDHCI_CDNS_HRS06_TUNE, val)))
 		return -EINVAL;
@@ -178,11 +179,23 @@ static int sdhci_cdns_set_tune_val(struct sdhci_cdns_plat *plat,
 	tmp = readl(reg);
 	tmp &= ~SDHCI_CDNS_HRS06_TUNE;
 	tmp |= FIELD_PREP(SDHCI_CDNS_HRS06_TUNE, val);
-	tmp |= SDHCI_CDNS_HRS06_TUNE_UP;
-	writel(tmp, reg);
 
-	return readl_poll_timeout(reg, tmp, !(tmp & SDHCI_CDNS_HRS06_TUNE_UP),
-				  1);
+	/*
+	 * Workaround for IP errata:
+	 * The IP6116 SD/eMMC PHY design has a timing issue on receive data
+	 * path. Send tune request twice.
+	 */
+	for (i = 0; i < 2; i++) {
+		tmp |= SDHCI_CDNS_HRS06_TUNE_UP;
+		writel(tmp, reg);
+
+		ret = readl_poll_timeout(reg, tmp,
+					 !(tmp & SDHCI_CDNS_HRS06_TUNE_UP), 1);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
 }
 
 static int __maybe_unused sdhci_cdns_execute_tuning(struct udevice *dev,
diff --git a/drivers/mmc/sdhci.c b/drivers/mmc/sdhci.c
index 01fa5a9d4d..9b7c5f8f68 100644
--- a/drivers/mmc/sdhci.c
+++ b/drivers/mmc/sdhci.c
@@ -15,12 +15,7 @@
 #include <mmc.h>
 #include <sdhci.h>
 #include <dm.h>
-
-#if defined(CONFIG_FIXED_SDHCI_ALIGNED_BUFFER)
-void *aligned_buffer = (void *)CONFIG_FIXED_SDHCI_ALIGNED_BUFFER;
-#else
-void *aligned_buffer;
-#endif
+#include <linux/dma-mapping.h>
 
 static void sdhci_reset(struct sdhci_host *host, u8 mask)
 {
@@ -71,8 +66,8 @@ static void sdhci_transfer_pio(struct sdhci_host *host, struct mmc_data *data)
 }
 
 #if CONFIG_IS_ENABLED(MMC_SDHCI_ADMA)
-static void sdhci_adma_desc(struct sdhci_host *host, char *buf, u16 len,
-			    bool end)
+static void sdhci_adma_desc(struct sdhci_host *host, dma_addr_t dma_addr,
+			    u16 len, bool end)
 {
 	struct sdhci_adma_desc *desc;
 	u8 attr;
@@ -88,9 +83,9 @@ static void sdhci_adma_desc(struct sdhci_host *host, char *buf, u16 len,
 	desc->attr = attr;
 	desc->len = len;
 	desc->reserved = 0;
-	desc->addr_lo = (dma_addr_t)buf;
+	desc->addr_lo = lower_32_bits(dma_addr);
 #ifdef CONFIG_DMA_ADDR_T_64BIT
-	desc->addr_hi = (u64)buf >> 32;
+	desc->addr_hi = upper_32_bits(dma_addr);
 #endif
 }
 
@@ -100,22 +95,17 @@ static void sdhci_prepare_adma_table(struct sdhci_host *host,
 	uint trans_bytes = data->blocksize * data->blocks;
 	uint desc_count = DIV_ROUND_UP(trans_bytes, ADMA_MAX_LEN);
 	int i = desc_count;
-	char *buf;
+	dma_addr_t dma_addr = host->start_addr;
 
 	host->desc_slot = 0;
 
-	if (data->flags & MMC_DATA_READ)
-		buf = data->dest;
-	else
-		buf = (char *)data->src;
-
 	while (--i) {
-		sdhci_adma_desc(host, buf, ADMA_MAX_LEN, false);
-		buf += ADMA_MAX_LEN;
+		sdhci_adma_desc(host, dma_addr, ADMA_MAX_LEN, false);
+		dma_addr += ADMA_MAX_LEN;
 		trans_bytes -= ADMA_MAX_LEN;
 	}
 
-	sdhci_adma_desc(host, buf, trans_bytes, true);
+	sdhci_adma_desc(host, dma_addr, trans_bytes, true);
 
 	flush_cache((dma_addr_t)host->adma_desc_table,
 		    ROUND(desc_count * sizeof(struct sdhci_adma_desc),
@@ -131,11 +121,12 @@ static void sdhci_prepare_dma(struct sdhci_host *host, struct mmc_data *data,
 			      int *is_aligned, int trans_bytes)
 {
 	unsigned char ctrl;
+	void *buf;
 
 	if (data->flags == MMC_DATA_READ)
-		host->start_addr = (dma_addr_t)data->dest;
+		buf = data->dest;
 	else
-		host->start_addr = (dma_addr_t)data->src;
+		buf = (void *)data->src;
 
 	ctrl = sdhci_readb(host, SDHCI_HOST_CONTROL);
 	ctrl &= ~SDHCI_CTRL_DMA_MASK;
@@ -145,37 +136,30 @@ static void sdhci_prepare_dma(struct sdhci_host *host, struct mmc_data *data,
 		ctrl |= SDHCI_CTRL_ADMA32;
 	sdhci_writeb(host, ctrl, SDHCI_HOST_CONTROL);
 
-	if (host->flags & USE_SDMA) {
-		if ((host->quirks & SDHCI_QUIRK_32BIT_DMA_ADDR) &&
-		    (host->start_addr & 0x7) != 0x0) {
-			*is_aligned = 0;
-			host->start_addr = (unsigned long)aligned_buffer;
-			if (data->flags != MMC_DATA_READ)
-				memcpy(aligned_buffer, data->src, trans_bytes);
-		}
-
-#if defined(CONFIG_FIXED_SDHCI_ALIGNED_BUFFER)
-		/*
-		 * Always use this bounce-buffer when
-		 * CONFIG_FIXED_SDHCI_ALIGNED_BUFFER is defined
-		 */
+	if (host->flags & USE_SDMA &&
+	    (host->force_align_buffer ||
+	     (host->quirks & SDHCI_QUIRK_32BIT_DMA_ADDR &&
+	      ((unsigned long)buf & 0x7) != 0x0))) {
 		*is_aligned = 0;
-		host->start_addr = (unsigned long)aligned_buffer;
 		if (data->flags != MMC_DATA_READ)
-			memcpy(aligned_buffer, data->src, trans_bytes);
-#endif
-		sdhci_writel(host, host->start_addr, SDHCI_DMA_ADDRESS);
+			memcpy(host->align_buffer, buf, trans_bytes);
+		buf = host->align_buffer;
+	}
 
+	host->start_addr = dma_map_single(buf, trans_bytes,
+					  mmc_get_dma_dir(data));
+
+	if (host->flags & USE_SDMA) {
+		sdhci_writel(host, host->start_addr, SDHCI_DMA_ADDRESS);
 	} else if (host->flags & (USE_ADMA | USE_ADMA64)) {
 		sdhci_prepare_adma_table(host, data);
 
-		sdhci_writel(host, (u32)host->adma_addr, SDHCI_ADMA_ADDRESS);
+		sdhci_writel(host, lower_32_bits(host->adma_addr),
+			     SDHCI_ADMA_ADDRESS);
 		if (host->flags & USE_ADMA64)
-			sdhci_writel(host, (u64)host->adma_addr >> 32,
+			sdhci_writel(host, upper_32_bits(host->adma_addr),
 				     SDHCI_ADMA_ADDRESS_HI);
 	}
-
-	flush_cache(host->start_addr, ROUND(trans_bytes, ARCH_DMA_MINALIGN));
 }
 #else
 static void sdhci_prepare_dma(struct sdhci_host *host, struct mmc_data *data,
@@ -231,6 +215,10 @@ static int sdhci_transfer_data(struct sdhci_host *host, struct mmc_data *data)
 			return -ETIMEDOUT;
 		}
 	} while (!(stat & SDHCI_INT_DATA_END));
+
+	dma_unmap_single(host->start_addr, data->blocks * data->blocksize,
+			 mmc_get_dma_dir(data));
+
 	return 0;
 }
 
@@ -381,7 +369,7 @@ static int sdhci_send_command(struct mmc *mmc, struct mmc_cmd *cmd,
 	if (!ret) {
 		if ((host->quirks & SDHCI_QUIRK_32BIT_DMA_ADDR) &&
 				!is_aligned && (data->flags == MMC_DATA_READ))
-			memcpy(data->dest, aligned_buffer, trans_bytes);
+			memcpy(data->dest, host->align_buffer, trans_bytes);
 		return 0;
 	}
 
@@ -537,7 +525,7 @@ static void sdhci_set_power(struct sdhci_host *host, unsigned short power)
 
 void sdhci_set_uhs_timing(struct sdhci_host *host)
 {
-	struct mmc *mmc = (struct mmc *)host->mmc;
+	struct mmc *mmc = host->mmc;
 	u32 reg;
 
 	reg = sdhci_readw(host, SDHCI_HOST_CONTROL2);
@@ -630,14 +618,23 @@ static int sdhci_init(struct mmc *mmc)
 
 	sdhci_reset(host, SDHCI_RESET_ALL);
 
-	if ((host->quirks & SDHCI_QUIRK_32BIT_DMA_ADDR) && !aligned_buffer) {
-		aligned_buffer = memalign(8, 512*1024);
-		if (!aligned_buffer) {
+#if defined(CONFIG_FIXED_SDHCI_ALIGNED_BUFFER)
+	host->align_buffer = (void *)CONFIG_FIXED_SDHCI_ALIGNED_BUFFER;
+	/*
+	 * Always use this bounce-buffer when CONFIG_FIXED_SDHCI_ALIGNED_BUFFER
+	 * is defined.
+	 */
+	host->force_align_buffer = true;
+#else
+	if (host->quirks & SDHCI_QUIRK_32BIT_DMA_ADDR) {
+		host->align_buffer = memalign(8, 512 * 1024);
+		if (!host->align_buffer) {
 			printf("%s: Aligned buffer alloc failed!!!\n",
 			       __func__);
 			return -ENOMEM;
 		}
 	}
+#endif
 
 	sdhci_set_power(host, fls(mmc->cfg->voltages) - 1);
 
@@ -741,8 +738,7 @@ int sdhci_setup_cfg(struct mmc_config *cfg, struct sdhci_host *host,
 		       __func__);
 		return -EINVAL;
 	}
-	host->adma_desc_table = (struct sdhci_adma_desc *)
-				memalign(ARCH_DMA_MINALIGN, ADMA_TABLE_SZ);
+	host->adma_desc_table = memalign(ARCH_DMA_MINALIGN, ADMA_TABLE_SZ);
 
 	host->adma_addr = (dma_addr_t)host->adma_desc_table;
 #ifdef CONFIG_DMA_ADDR_T_64BIT
diff --git a/drivers/mmc/tmio-common.c b/drivers/mmc/tmio-common.c
index 092b740f56..faf18191b3 100644
--- a/drivers/mmc/tmio-common.c
+++ b/drivers/mmc/tmio-common.c
@@ -4,7 +4,6 @@
  *   Author: Masahiro Yamada <yamada.masahiro@socionext.com>
  */
 
-#include <asm/dma-mapping.h>
 #include <common.h>
 #include <clk.h>
 #include <cpu_func.h>
@@ -14,7 +13,7 @@
 #include <dm/device_compat.h>
 #include <dm/pinctrl.h>
 #include <linux/compat.h>
-#include <linux/dma-direction.h>
+#include <linux/dma-mapping.h>
 #include <linux/io.h>
 #include <linux/sizes.h>
 #include <power/regulator.h>
@@ -353,7 +352,7 @@ static int tmio_sd_dma_xfer(struct udevice *dev, struct mmc_data *data)
 	if (poll_flag == TMIO_SD_DMA_INFO1_END_RD)
 		udelay(1);
 
-	dma_unmap_single(buf, len, dir);
+	dma_unmap_single(dma_addr, len, dir);
 
 	return ret;
 }
diff --git a/drivers/mtd/nand/raw/denali.c b/drivers/mtd/nand/raw/denali.c
index f51d3e25c7..b525b1be54 100644
--- a/drivers/mtd/nand/raw/denali.c
+++ b/drivers/mtd/nand/raw/denali.c
@@ -5,14 +5,13 @@
  * Copyright (C) 2009-2010, Intel Corporation and its suppliers.
  */
 
-#include <asm/dma-mapping.h>
 #include <dm.h>
 #include <malloc.h>
 #include <nand.h>
 #include <dm/device_compat.h>
 #include <dm/devres.h>
 #include <linux/bitfield.h>
-#include <linux/dma-direction.h>
+#include <linux/dma-mapping.h>
 #include <linux/err.h>
 #include <linux/errno.h>
 #include <linux/io.h>
@@ -581,7 +580,7 @@ static int denali_dma_xfer(struct denali_nand_info *denali, void *buf,
 
 	iowrite32(0, denali->reg + DMA_ENABLE);
 
-	dma_unmap_single(buf, size, dir);
+	dma_unmap_single(dma_addr, size, dir);
 
 	if (irq_status & INTR__ERASED_PAGE)
 		memset(buf, 0xff, size);
diff --git a/drivers/net/altera_tse.c b/drivers/net/altera_tse.c
index aabddd6bb6..a511068d9a 100644
--- a/drivers/net/altera_tse.c
+++ b/drivers/net/altera_tse.c
@@ -17,7 +17,7 @@
 #include <miiphy.h>
 #include <net.h>
 #include <asm/cache.h>
-#include <asm/dma-mapping.h>
+#include <linux/dma-mapping.h>
 #include <asm/io.h>
 #include "altera_tse.h"
 
diff --git a/drivers/net/ftmac110.c b/drivers/net/ftmac110.c
index 1fa93d4d58..84082407d2 100644
--- a/drivers/net/ftmac110.c
+++ b/drivers/net/ftmac110.c
@@ -12,7 +12,7 @@
 #include <net.h>
 #include <linux/errno.h>
 #include <asm/io.h>
-#include <asm/dma-mapping.h>
+#include <linux/dma-mapping.h>
 
 #if defined(CONFIG_MII) || defined(CONFIG_CMD_MII)
 #include <miiphy.h>
diff --git a/drivers/net/macb.c b/drivers/net/macb.c
index 0d4929bec1..631b53b093 100644
--- a/drivers/net/macb.c
+++ b/drivers/net/macb.c
@@ -38,7 +38,7 @@
 
 #include <linux/mii.h>
 #include <asm/io.h>
-#include <asm/dma-mapping.h>
+#include <linux/dma-mapping.h>
 #include <asm/arch/clk.h>
 #include <linux/errno.h>
 
@@ -342,7 +342,7 @@ static int _macb_send(struct macb_device *macb, const char *name, void *packet,
 		udelay(1);
 	}
 
-	dma_unmap_single(packet, length, DMA_TO_DEVICE);
+	dma_unmap_single(paddr, length, DMA_TO_DEVICE);
 
 	if (i <= MACB_TX_TIMEOUT) {
 		if (ctrl & MACB_BIT(TX_UNDERRUN))
diff --git a/drivers/soc/ti/k3-navss-ringacc.c b/drivers/soc/ti/k3-navss-ringacc.c
index 8cbfe2bf49..c5661c52fa 100644
--- a/drivers/soc/ti/k3-navss-ringacc.c
+++ b/drivers/soc/ti/k3-navss-ringacc.c
@@ -9,7 +9,6 @@
 #include <cpu_func.h>
 #include <asm/io.h>
 #include <malloc.h>
-#include <asm/dma-mapping.h>
 #include <asm/bitops.h>
 #include <dm.h>
 #include <dm/device_compat.h>
@@ -17,6 +16,7 @@
 #include <dm/read.h>
 #include <dm/uclass.h>
 #include <linux/compat.h>
+#include <linux/dma-mapping.h>
 #include <linux/err.h>
 #include <linux/soc/ti/k3-navss-ringacc.h>
 #include <linux/soc/ti/ti_sci_protocol.h>
diff --git a/drivers/ufs/ufs.c b/drivers/ufs/ufs.c
index c9346c2edc..24e1bc2a19 100644
--- a/drivers/ufs/ufs.c
+++ b/drivers/ufs/ufs.c
@@ -19,7 +19,7 @@
 #include <hexdump.h>
 #include <scsi.h>
 
-#include <asm/dma-mapping.h>
+#include <linux/dma-mapping.h>
 
 #include "ufs.h"
 
diff --git a/drivers/usb/cdns3/gadget.c b/drivers/usb/cdns3/gadget.c
index 22e90a5717..8377eb458b 100644
--- a/drivers/usb/cdns3/gadget.c
+++ b/drivers/usb/cdns3/gadget.c
@@ -63,7 +63,7 @@
 #include <linux/usb/gadget.h>
 #include <linux/compat.h>
 #include <linux/iopoll.h>
-#include <asm/dma-mapping.h>
+#include <linux/dma-mapping.h>
 #include <linux/bitmap.h>
 #include <linux/bug.h>
 
diff --git a/drivers/usb/dwc3/core.c b/drivers/usb/dwc3/core.c
index c5066529b7..4ec3f6df6a 100644
--- a/drivers/usb/dwc3/core.c
+++ b/drivers/usb/dwc3/core.c
@@ -17,9 +17,9 @@
 #include <cpu_func.h>
 #include <malloc.h>
 #include <dwc3-uboot.h>
-#include <asm/dma-mapping.h>
 #include <dm/device_compat.h>
 #include <dm/devres.h>
+#include <linux/dma-mapping.h>
 #include <linux/err.h>
 #include <linux/ioport.h>
 #include <dm.h>
@@ -288,8 +288,8 @@ static int dwc3_setup_scratch_buffers(struct dwc3 *dwc)
 	return 0;
 
 err1:
-	dma_unmap_single((void *)(uintptr_t)dwc->scratch_addr, dwc->nr_scratch *
-			 DWC3_SCRATCHBUF_SIZE, DMA_BIDIRECTIONAL);
+	dma_unmap_single(scratch_addr, dwc->nr_scratch * DWC3_SCRATCHBUF_SIZE,
+			 DMA_BIDIRECTIONAL);
 
 err0:
 	return ret;
@@ -303,7 +303,7 @@ static void dwc3_free_scratch_buffers(struct dwc3 *dwc)
 	if (!dwc->nr_scratch)
 		return;
 
-	dma_unmap_single((void *)(uintptr_t)dwc->scratch_addr, dwc->nr_scratch *
+	dma_unmap_single(dwc->scratch_addr, dwc->nr_scratch *
 			 DWC3_SCRATCHBUF_SIZE, DMA_BIDIRECTIONAL);
 	kfree(dwc->scratchbuf);
 }
diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c
index 1502d67362..e445c705f1 100644
--- a/drivers/usb/dwc3/gadget.c
+++ b/drivers/usb/dwc3/gadget.c
@@ -16,10 +16,10 @@
 #include <common.h>
 #include <cpu_func.h>
 #include <malloc.h>
-#include <asm/dma-mapping.h>
 #include <dm/device_compat.h>
 #include <dm/devres.h>
 #include <linux/bug.h>
+#include <linux/dma-mapping.h>
 #include <linux/list.h>
 
 #include <linux/usb/ch9.h>
diff --git a/drivers/usb/gadget/udc/udc-core.c b/drivers/usb/gadget/udc/udc-core.c
index a33ab5c95d..7f73926cb3 100644
--- a/drivers/usb/gadget/udc/udc-core.c
+++ b/drivers/usb/gadget/udc/udc-core.c
@@ -18,7 +18,7 @@
 #include <linux/compat.h>
 #include <malloc.h>
 #include <asm/cache.h>
-#include <asm/dma-mapping.h>
+#include <linux/dma-mapping.h>
 #include <common.h>
 #include <dm.h>
 #include <dm/device-internal.h>
@@ -67,7 +67,7 @@ void usb_gadget_unmap_request(struct usb_gadget *gadget,
 	if (req->length == 0)
 		return;
 
-	dma_unmap_single((void *)(uintptr_t)req->dma, req->length,
+	dma_unmap_single(req->dma, req->length,
 			 is_in ? DMA_TO_DEVICE : DMA_FROM_DEVICE);
 }
 EXPORT_SYMBOL_GPL(usb_gadget_unmap_request);
diff --git a/include/linux/dma-mapping.h b/include/linux/dma-mapping.h
new file mode 100644
index 0000000000..20b6d60dd8
--- /dev/null
+++ b/include/linux/dma-mapping.h
@@ -0,0 +1,63 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _LINUX_DMA_MAPPING_H
+#define _LINUX_DMA_MAPPING_H
+
+#include <linux/dma-direction.h>
+#include <linux/types.h>
+#include <asm/dma-mapping.h>
+#include <cpu_func.h>
+
+#define dma_mapping_error(x, y)	0
+
+/**
+ * Map a buffer to make it available to the DMA device
+ *
+ * Linux-like DMA API that is intended to be used from drivers. This hides the
+ * underlying cache operation from drivers. Call this before starting the DMA
+ * transfer. In most of architectures in U-Boot, the virtual address matches to
+ * the physical address (but we have exceptions like sandbox). U-Boot does not
+ * support iommu at the driver level, so it also matches to the DMA address.
+ * Hence, this helper currently just performs the cache operation, then returns
+ * straight-mapped dma_address, which is intended to be set to the register of
+ * the DMA device.
+ *
+ * @vaddr: address of the buffer
+ * @len: length of the buffer
+ * @dir: the direction of DMA
+ */
+static inline dma_addr_t dma_map_single(void *vaddr, size_t len,
+					enum dma_data_direction dir)
+{
+	unsigned long addr = (unsigned long)vaddr;
+
+	len = ALIGN(len, ARCH_DMA_MINALIGN);
+
+	if (dir == DMA_FROM_DEVICE)
+		invalidate_dcache_range(addr, addr + len);
+	else
+		flush_dcache_range(addr, addr + len);
+
+	return addr;
+}
+
+/**
+ * Unmap a buffer to make it available to CPU
+ *
+ * Linux-like DMA API that is intended to be used from drivers. This hides the
+ * underlying cache operation from drivers. Call this after finishin the DMA
+ * transfer.
+ *
+ * @addr: DMA address
+ * @len: length of the buffer
+ * @dir: the direction of DMA
+ */
+static inline void dma_unmap_single(dma_addr_t addr, size_t len,
+				    enum dma_data_direction dir)
+{
+	len = ALIGN(len, ARCH_DMA_MINALIGN);
+
+	if (dir != DMA_TO_DEVICE)
+		invalidate_dcache_range(addr, addr + len);
+}
+
+#endif
diff --git a/include/mmc.h b/include/mmc.h
index b5cb514f57..71e2e1735a 100644
--- a/include/mmc.h
+++ b/include/mmc.h
@@ -12,6 +12,7 @@
 #include <linux/list.h>
 #include <linux/sizes.h>
 #include <linux/compiler.h>
+#include <linux/dma-direction.h>
 #include <part.h>
 
 #if CONFIG_IS_ENABLED(MMC_HS200_SUPPORT)
@@ -880,4 +881,9 @@ int mmc_get_env_dev(void);
  */
 struct blk_desc *mmc_get_blk_desc(struct mmc *mmc);
 
+static inline enum dma_data_direction mmc_get_dma_dir(struct mmc_data *data)
+{
+	return data->flags & MMC_DATA_WRITE ? DMA_TO_DEVICE : DMA_FROM_DEVICE;
+}
+
 #endif /* _MMC_H_ */
diff --git a/include/sdhci.h b/include/sdhci.h
index 01addb7a60..7f8feefa45 100644
--- a/include/sdhci.h
+++ b/include/sdhci.h
@@ -9,6 +9,7 @@
 #ifndef __SDHCI_HW_H
 #define __SDHCI_HW_H
 
+#include <linux/types.h>
 #include <asm/io.h>
 #include <mmc.h>
 #include <asm/gpio.h>
@@ -321,6 +322,8 @@ struct sdhci_host {
 	uint	voltages;
 
 	struct mmc_config cfg;
+	void *align_buffer;
+	bool force_align_buffer;
 	dma_addr_t start_addr;
 	int flags;
 #define USE_SDMA	(0x1 << 0)