[PATCH] mmc: sdhci-cadence: Add support for Cadence sdmmc v6

KuanLim.Lee kuanlim.lee at starfivetech.com
Tue Nov 7 07:23:48 CET 2023


Hi Jaehoon,

On 11/1/23 08:20, Jaehoon Chung wrote:
> From: Jaehoon Chung <jh80.chung at samsung.com>
> Hi
> 
> On 10/3/23 16:22, Kuan Lim Lee wrote:
> > From: Kuan Lim Lee <kuanlim.lee at linux.starfivetech.com>
> >
> > Cadence SDMMC v6 controller has a lot of changes on initialize
> > compared to v4 controller. PHY is needed by v6 controller.
> >
> > Signed-off-by: Kuan Lim Lee <kuanlim.lee at starfivetech.com>
> > Reviewed-by: Alex Soo <yuklin.soo at starfivetech.com>
> > Reviewed-by: Wei Liang Lim <weiliang.lim at starfivetech.com>
> 
> I didn't see their Reviewed-by tag in mailing list.
They are my colleagues who collaborated develop code with me.
I think I should them in Signed-off-by, and put your name in Reviewed-by

> 
> > ---
> >  drivers/mmc/Kconfig              |  13 ++
> >  drivers/mmc/Makefile             |   1 +
> >  drivers/mmc/sdhci-cadence.c      |  63 ++-----
> >  drivers/mmc/sdhci-cadence.h      |  68 +++++++
> >  drivers/mmc/sdhci-cadence6-phy.c | 302
> > +++++++++++++++++++++++++++++++
> >  5 files changed, 396 insertions(+), 51 deletions(-)  create mode
> > 100644 drivers/mmc/sdhci-cadence.h  create mode 100644
> > drivers/mmc/sdhci-cadence6-phy.c
> >
> > diff --git a/drivers/mmc/Kconfig b/drivers/mmc/Kconfig index
> > de01b9687b..cec881d862 100644
> > --- a/drivers/mmc/Kconfig
> > +++ b/drivers/mmc/Kconfig
> > @@ -573,6 +573,19 @@ config MMC_SDHCI_CADENCE
> >
> >  	  If unsure, say N.
> >
> > +config MMC_SDHCI_CADENCE_V6
> > +	bool "SDHCI support for the Cadence SD/SDIO/eMMC controller &
> driver version 6"
> > +	depends on BLK && DM_MMC
> > +	depends on MMC_SDHCI
> > +	depends on OF_CONTROL
> > +	select MMC_SDHCI_CADENCE
> > +	help
> > +	  This selects the Cadence SD/SDIO/eMMC driver version 6.
> > +
> > +	  If you have a controller with this interface, say Y here.
> > +
> > +	  If unsure, say N.
> > +
> >  config MMC_SDHCI_AM654
> >  	bool "SDHCI Controller on TI's Am654 devices"
> >  	depends on ARCH_K3
> > diff --git a/drivers/mmc/Makefile b/drivers/mmc/Makefile index
> > 2c65c4765a..cdcce55b8b 100644
> > --- a/drivers/mmc/Makefile
> > +++ b/drivers/mmc/Makefile
> > @@ -61,6 +61,7 @@ obj-$(CONFIG_MMC_SDHCI_ATMEL)		+=
> atmel_sdhci.o
> >  obj-$(CONFIG_MMC_SDHCI_BCM2835)		+= bcm2835_sdhci.o
> >  obj-$(CONFIG_MMC_SDHCI_BCMSTB)		+= bcmstb_sdhci.o
> >  obj-$(CONFIG_MMC_SDHCI_CADENCE)		+= sdhci-cadence.o
> > +obj-$(CONFIG_MMC_SDHCI_CADENCE_V6)	+= sdhci-cadence6-phy.o
> >  obj-$(CONFIG_MMC_SDHCI_AM654)		+= am654_sdhci.o
> >  obj-$(CONFIG_MMC_SDHCI_IPROC)		+= iproc_sdhci.o
> >  obj-$(CONFIG_MMC_SDHCI_KONA)		+= kona_sdhci.o
> > diff --git a/drivers/mmc/sdhci-cadence.c b/drivers/mmc/sdhci-cadence.c
> > index 327a05ad11..d7a270e74c 100644
> > --- a/drivers/mmc/sdhci-cadence.c
> > +++ b/drivers/mmc/sdhci-cadence.c
> > @@ -17,56 +17,7 @@
> >  #include <linux/libfdt.h>
> >  #include <mmc.h>
> >  #include <sdhci.h>
> > -
> > -/* HRS - Host Register Set (specific to Cadence) */
> > -#define SDHCI_CDNS_HRS04		0x10		/* PHY access port */
> > -#define   SDHCI_CDNS_HRS04_ACK			BIT(26)
> > -#define   SDHCI_CDNS_HRS04_RD			BIT(25)
> > -#define   SDHCI_CDNS_HRS04_WR			BIT(24)
> > -#define   SDHCI_CDNS_HRS04_RDATA		GENMASK(23, 16)
> > -#define   SDHCI_CDNS_HRS04_WDATA		GENMASK(15, 8)
> > -#define   SDHCI_CDNS_HRS04_ADDR			GENMASK(5, 0)
> > -
> > -#define SDHCI_CDNS_HRS06		0x18		/* eMMC control */
> > -#define   SDHCI_CDNS_HRS06_TUNE_UP		BIT(15)
> > -#define   SDHCI_CDNS_HRS06_TUNE			GENMASK(13, 8)
> > -#define   SDHCI_CDNS_HRS06_MODE			GENMASK(2, 0)
> > -#define   SDHCI_CDNS_HRS06_MODE_SD		0x0
> > -#define   SDHCI_CDNS_HRS06_MODE_MMC_SDR		0x2
> > -#define   SDHCI_CDNS_HRS06_MODE_MMC_DDR		0x3
> > -#define   SDHCI_CDNS_HRS06_MODE_MMC_HS200	0x4
> > -#define   SDHCI_CDNS_HRS06_MODE_MMC_HS400	0x5
> > -#define   SDHCI_CDNS_HRS06_MODE_MMC_HS400ES	0x6
> > -
> > -/* SRS - Slot Register Set (SDHCI-compatible) */
> > -#define SDHCI_CDNS_SRS_BASE		0x200
> > -
> > -/* PHY */
> > -#define SDHCI_CDNS_PHY_DLY_SD_HS	0x00
> > -#define SDHCI_CDNS_PHY_DLY_SD_DEFAULT	0x01
> > -#define SDHCI_CDNS_PHY_DLY_UHS_SDR12	0x02
> > -#define SDHCI_CDNS_PHY_DLY_UHS_SDR25	0x03
> > -#define SDHCI_CDNS_PHY_DLY_UHS_SDR50	0x04
> > -#define SDHCI_CDNS_PHY_DLY_UHS_DDR50	0x05
> > -#define SDHCI_CDNS_PHY_DLY_EMMC_LEGACY	0x06
> > -#define SDHCI_CDNS_PHY_DLY_EMMC_SDR	0x07
> > -#define SDHCI_CDNS_PHY_DLY_EMMC_DDR	0x08
> > -#define SDHCI_CDNS_PHY_DLY_SDCLK	0x0b
> > -#define SDHCI_CDNS_PHY_DLY_HSMMC	0x0c
> > -#define SDHCI_CDNS_PHY_DLY_STROBE	0x0d
> > -
> > -/*
> > - * The tuned val register is 6 bit-wide, but not the whole of the
> > range is
> > - * available.  The range 0-42 seems to be available (then 43 wraps
> > around to 0)
> > - * but I am not quite sure if it is official.  Use only 0 to 39 for safety.
> > - */
> > -#define SDHCI_CDNS_MAX_TUNING_LOOP	40
> > -
> > -struct sdhci_cdns_plat {
> > -	struct mmc_config cfg;
> > -	struct mmc mmc;
> > -	void __iomem *hrs_addr;
> > -};
> > +#include "sdhci-cadence.h"
> >
> >  struct sdhci_cdns_phy_cfg {
> >  	const char *property;
> > @@ -112,8 +63,11 @@ static int sdhci_cdns_write_phy_reg(struct
> > sdhci_cdns_plat *plat,  }
> >
> >  static int sdhci_cdns_phy_init(struct sdhci_cdns_plat *plat,
> > -				const void *fdt, int nodeoffset)
> > +			       const void *fdt, int nodeoffset)
> 
> Is there any reason to touch this?
I thought the space key is misaligned. I think I better not to touch it.
> 
> 
> >  {
> > +	if (IS_ENABLED(CONFIG_MMC_SDHCI_CADENCE_V6))
> > +		return sdhci_cdns6_phy_init(plat,
> SDHCI_CDNS_HRS06_MODE_SD);
> > +
> >  	const fdt32_t *prop;
> >  	int ret, i;
> >
> > @@ -163,6 +117,9 @@ static void sdhci_cdns_set_control_reg(struct
> sdhci_host *host)
> >  	tmp &= ~SDHCI_CDNS_HRS06_MODE;
> >  	tmp |= FIELD_PREP(SDHCI_CDNS_HRS06_MODE, mode);
> >  	writel(tmp, plat->hrs_addr + SDHCI_CDNS_HRS06);
> > +
> > +	if (IS_ENABLED(CONFIG_MMC_SDHCI_CADENCE_V6))
> > +		sdhci_cdns6_phy_init(plat, mode);
> >  }
> >
> >  static const struct sdhci_ops sdhci_cdns_ops = { @@ -172,6 +129,9 @@
> > static const struct sdhci_ops sdhci_cdns_ops = {  static int
> > sdhci_cdns_set_tune_val(struct sdhci_cdns_plat *plat,
> >  				   unsigned int val)
> >  {
> > +	if (IS_ENABLED(CONFIG_MMC_SDHCI_CADENCE_V6))
> > +		return sdhci_cdns6_set_tune_val(plat, val);
> > +
> >  	void __iomem *reg = plat->hrs_addr + SDHCI_CDNS_HRS06;
> >  	u32 tmp;
> >  	int i, ret;
> > @@ -301,6 +261,7 @@ static int sdhci_cdns_probe(struct udevice *dev)
> > static const struct udevice_id sdhci_cdns_match[] = {
> >  	{ .compatible = "socionext,uniphier-sd4hc" },
> >  	{ .compatible = "cdns,sd4hc" },
> > +	{ .compatible = "cdns,sd6hc" },
> >  	{ /* sentinel */ }
> >  };
> >
> > diff --git a/drivers/mmc/sdhci-cadence.h b/drivers/mmc/sdhci-cadence.h
> > new file mode 100644 index 0000000000..2c42cff2f3
> > --- /dev/null
> > +++ b/drivers/mmc/sdhci-cadence.h
> > @@ -0,0 +1,68 @@
> > +/* SPDX-License-Identifier: GPL-2.0-only */
> > +/*
> > + * Copyright (C) 2022 Starfive.
> > + *   Author: Kuan Lim Lee <kuanlim.lee at starfivetech.com>
> 
> 
> Does it right Copyright and author? the below codes are taken from sdhci-
> cadence.c.
Some parts of #define are taken from sdchi-cadence.c, some parts of #define
are written by me. I think I still better put back the original Copyright and 
author. 

> 
> > + */
> > +
> > +#ifndef SDHCI_CADENCE_H_
> > +#define SDHCI_CADENCE_H_
> > +
> > +/* HRS - Host Register Set (specific to Cadence) */
> > +/* PHY access port */
> > +#define SDHCI_CDNS_HRS04		0x10
> > +/* Cadence V4 HRS04 Description*/
> > +#define   SDHCI_CDNS_HRS04_ACK			BIT(26)
> > +#define   SDHCI_CDNS_HRS04_RD			BIT(25)
> > +#define   SDHCI_CDNS_HRS04_WR			BIT(24)
> > +#define   SDHCI_CDNS_HRS04_RDATA		GENMASK(23, 16)
> > +#define   SDHCI_CDNS_HRS04_WDATA		GENMASK(15, 8)
> > +#define   SDHCI_CDNS_HRS04_ADDR			GENMASK(5, 0)
> > +
> > +#define SDHCI_CDNS_HRS05		0x14
> > +
> > +/* eMMC control */
> > +#define SDHCI_CDNS_HRS06		0x18
> > +#define   SDHCI_CDNS_HRS06_TUNE_UP		BIT(15)
> > +#define   SDHCI_CDNS_HRS06_TUNE			GENMASK(13, 8)
> > +#define   SDHCI_CDNS_HRS06_MODE			GENMASK(2,
> 0)
> > +#define   SDHCI_CDNS_HRS06_MODE_SD		0x0
> > +#define   SDHCI_CDNS_HRS06_MODE_MMC_SDR		0x2
> > +#define   SDHCI_CDNS_HRS06_MODE_MMC_DDR		0x3
> > +#define   SDHCI_CDNS_HRS06_MODE_MMC_HS200	0x4
> > +#define   SDHCI_CDNS_HRS06_MODE_MMC_HS400	0x5
> > +#define   SDHCI_CDNS_HRS06_MODE_MMC_HS400ES	0x6
> > +
> > +/* SRS - Slot Register Set (SDHCI-compatible) */
> > +#define SDHCI_CDNS_SRS_BASE		0x200
> > +
> > +/* Cadence V4 PHY Setting*/
> > +#define SDHCI_CDNS_PHY_DLY_SD_HS	0x00
> > +#define SDHCI_CDNS_PHY_DLY_SD_DEFAULT	0x01
> > +#define SDHCI_CDNS_PHY_DLY_UHS_SDR12	0x02
> > +#define SDHCI_CDNS_PHY_DLY_UHS_SDR25	0x03
> > +#define SDHCI_CDNS_PHY_DLY_UHS_SDR50	0x04
> > +#define SDHCI_CDNS_PHY_DLY_UHS_DDR50	0x05
> > +#define SDHCI_CDNS_PHY_DLY_EMMC_LEGACY	0x06
> > +#define SDHCI_CDNS_PHY_DLY_EMMC_SDR	0x07
> > +#define SDHCI_CDNS_PHY_DLY_EMMC_DDR	0x08
> > +#define SDHCI_CDNS_PHY_DLY_SDCLK	0x0b
> > +#define SDHCI_CDNS_PHY_DLY_HSMMC	0x0c
> > +#define SDHCI_CDNS_PHY_DLY_STROBE	0x0d
> > +
> > +/*
> > + * The tuned val register is 6 bit-wide, but not the whole of the
> > +range is
> > + * available.  The range 0-42 seems to be available (then 43 wraps
> > +around to 0)
> > + * but I am not quite sure if it is official.  Use only 0 to 39 for safety.
> > + */
> > +#define SDHCI_CDNS_MAX_TUNING_LOOP	40
> > +
> > +struct sdhci_cdns_plat {
> > +	struct mmc_config cfg;
> > +	struct mmc mmc;
> > +	void __iomem *hrs_addr;
> > +};
> > +
> > +int sdhci_cdns6_phy_init(struct sdhci_cdns_plat *plat, u32 mode); int
> > +sdhci_cdns6_set_tune_val(struct sdhci_cdns_plat *plat, unsigned int
> > +val);
> > +
> > +#endif
> > diff --git a/drivers/mmc/sdhci-cadence6-phy.c
> > b/drivers/mmc/sdhci-cadence6-phy.c
> > new file mode 100644
> > index 0000000000..dd3df27dc8
> > --- /dev/null
> > +++ b/drivers/mmc/sdhci-cadence6-phy.c
> > @@ -0,0 +1,302 @@
> > +// SPDX-License-Identifier: GPL-2.0-or-platform_driver
> > +/*
> > + * Copyright (C) 2022 Starfive.
> 
> s/2022/2023?
I will put 2023.

> 
> > + *   Author: Kuan Lim Lee <kuanlim.lee at starfivetech.com>
> > + */
> > +
> > +#include <common.h>
> > +#include <dm.h>
> > +#include <asm/global_data.h>
> > +#include <dm/device_compat.h>
> > +#include <linux/bitfield.h>
> > +#include <linux/bitops.h>
> > +#include <linux/bug.h>
> > +#include <linux/io.h>
> > +#include <linux/iopoll.h>
> > +#include <linux/sizes.h>
> > +#include <linux/libfdt.h>
> > +#include <mmc.h>
> > +#include <sdhci.h>
> > +#include "sdhci-cadence.h"
> > +
> > +/* IO Delay Information */
> > +#define SDHCI_CDNS_HRS07		0X1C
> > +#define   SDHCI_CDNS_HRS07_RW_COMPENSATE		GENMASK(20,
> 16)
> > +#define   SDHCI_CDNS_HRS07_IDELAY_VAL			GENMASK(4,
> 0)
> > +
> > +#define SDHCI_CDNS_HRS09		0x24		/* PHY Control and
> Status */
> > +#define   SDHCI_CDNS_HRS09_RDDATA_EN		BIT(16)
> > +#define   SDHCI_CDNS_HRS09_RDCMD_EN		BIT(15)
> > +#define   SDHCI_CDNS_HRS09_EXTENDED_WR_MODE	BIT(3)
> > +#define   SDHCI_CDNS_HRS09_EXTENDED_RD_MODE	BIT(2)
> > +#define   SDHCI_CDNS_HRS09_PHY_INIT_COMPLETE	BIT(1)
> > +#define   SDHCI_CDNS_HRS09_PHY_SW_RESET		BIT(0)
> > +
> > +#define SDHCI_CDNS_HRS10		0x28		/* SDCLK adjustment
> */
> > +#define   SDHCI_CDNS_HRS10_HCSDCLKADJ		GENMASK(19, 16)
> > +
> > +#define SDHCI_CDNS_HRS16		0x40		/* CMD/DAT output
> delay */
> > +
> > +/* PHY Special Function Registers */
> > +//#define DLL_PHY_REG_BASE		0x2000
> > +
> > +/* register to control the DQ related timing */
> > +#define PHY_DQ_TIMING_REG_ADDR		0x2000
> > +
> > +/* register to control the DQS related timing */
> > +#define PHY_DQS_TIMING_REG_ADDR		0x2004
> > +
> > +/* register to control the gate and loopback control related timing */
> > +#define PHY_GATE_LPBK_CTRL_REG_ADDR	0x2008
> > +
> > +/* register to control the Master DLL logic */
> > +#define PHY_DLL_MASTER_CTRL_REG_ADDR	0x200C
> > +
> > +/* register to control the Slave DLL logic */
> > +#define PHY_DLL_SLAVE_CTRL_REG_ADDR	0x2010
> > +#define PHY_DLL_SLAVE_CTRL_REG_READ_DQS_CMD_DELAY
> 	GENMASK(31, 24)
> > +#define PHY_DLL_SLAVE_CTRL_REG_READ_DQS_DELAY
> 	GENMASK(7, 0)
> > +
> > +/* register to control the global settings for PHY */
> > +#define PHY_CTRL_REG_ADDR		0x2080
> > +
> > +struct phy_reg {
> > +	u32 phy_dqs_timing;
> > +	u32 phy_gate_lpbk_ctrl;
> > +	//u32 phy_dll_master_ctrl;
> 
> Remove unused code.
Noted.
> 
> > +	u32 phy_dll_slave_ctrl;
> > +	//u32 phy_ctrl;
> 
> ditto
Noted.
> 
> > +	u32 phy_dq_timing;
> > +	//cp_sw_half_cycle_shift; ASIC
> 
> Ditto
Noted.
> 
> > +};
> > +
> > +struct controller_reg {
> > +	u32 hrs07;
> > +	u32 hrs09;
> > +	u32 hrs10;
> > +	u32 hrs16;
> > +};
> 
> 
> As some reg values can be re-used. I don't know what its value means.
> e.g) 0x00380004 is sued sd_ds/emmc_sdr/emmc_ddr/emmc_hs200.
This value is generated by script provided by Cadence.
e.g.) phy_dqs_timing_reg (0x2004) is a 32 bits register, inside the register 
has few component: use_ext_lpbk_dqs, use_lpbk_dqs, use_phony_dqs,
use_phony_dqs_cmd, phony_dqs_sel, dqs_select_tsel_start and 
dqs_select_tsel_end.

Each components value is generated by script and I combined the value
in a 32 bits value, 0x00380004. 

> 
> I don't want to use a magic code. If you can add some macros, it's more
> readable.
Should I change the struct in way below, so that it is more reable?
static struct sdhci_cdns_phy_reg_pairs sd_ds_mode_setting[] = {
		{0x00380000, PHY_DQS_TIMING_REG_ADDR},
		{0x01A00040, PHY_GATE_LPBK_CTRL_REG_ADDR},
		{0x00000000, PHY_DLL_SLAVE_CTRL_REG_ADDR},
		{0x00000180, PHY_CTRL_REG_ADDR},
		{0x00000001, PHY_DQ_TIMING_REG_ADDR},
};

Or I have to put each value of components and put a convertion 
function in the code to change this to 32 bits value and program inti
register, phy_dqs_timing_reg (0x2004)?

> 
> > +
> > +static struct phy_reg sd_ds_phy_cfg = {
> > +	0x00380004,
> > +	0x01A00040,
> > +	0x00000000,
> > +	0x00000001,
> > +};
> > +
> > +static struct phy_reg emmc_sdr_phy_cfg = {
> > +	0x00380004,
> > +	0x01A00040,
> > +	0x00000000,
> > +	0x00000001,
> > +};
> > +
> > +static struct phy_reg emmc_ddr_phy_cfg = {
> > +	0x00380004,
> > +	0x01A00040,
> > +	0x00000000,
> > +	0x10000001,
> > +};
> > +
> > +static struct phy_reg emmc_hs200_phy_cfg = {
> > +	0x00380004,
> > +	0x01A00040,
> > +	0x00DADA00,
> > +	0x00000001,
> > +};
> > +
> > +static struct phy_reg emmc_hs400_phy_cfg = {
> > +	0x00280004,
> > +	0x01A00040,
> > +	0x00DAD800,
> > +	0x00000001,
> > +};
> > +
> > +static struct controller_reg sd_ds_ctrl_cfg = {
> > +	0x00080000,
> > +	0x0001800C,
> > +	0x00020000,
> > +	0x00000000,
> > +};
> > +
> > +static struct controller_reg emmc_sdr_ctrl_cfg = {
> > +	0x00080000,
> > +	0x0001800C,
> > +	0x00030000,
> > +	0x00000000,
> > +};
> > +
> > +static struct controller_reg emmc_ddr_ctrl_cfg = {
> > +	0x00090001,
> > +	0x0001800C,
> > +	0x00020000,
> > +	0x11000001,
> > +};
> > +
> > +static struct controller_reg emmc_hs200_ctrl_cfg = {
> > +	0x00090000,
> > +	0x00018000,
> > +	0x00080000,
> > +	0x00000000,
> > +};
> > +
> > +static struct controller_reg emmc_hs400_ctrl_cfg = {
> > +	0x00080000,
> > +	0x00018000,
> > +	0x00080000,
> > +	0x11000000,
> > +};
> > +
> > +static unsigned int sdhci_cdns6_read_phy_reg(struct sdhci_cdns_plat *plat,
> > +					     u32 addr)
> > +{
> > +	u32 data = 0;
> > +
> > +	writel(addr, plat->hrs_addr + SDHCI_CDNS_HRS04);
> > +	data = readl(plat->hrs_addr + SDHCI_CDNS_HRS05);
> > +	return data;
> > +}
> > +
> > +static void sdhci_cdns6_write_phy_reg(struct sdhci_cdns_plat *plat,
> > +				      u32 addr, u32 data)
> > +{
> > +	u32 readdat = 0;
> > +
> > +	writel(addr, plat->hrs_addr + SDHCI_CDNS_HRS04);
> > +	writel(data, plat->hrs_addr + SDHCI_CDNS_HRS05);
> > +	readdat = readl(plat->hrs_addr + SDHCI_CDNS_HRS05);
> > +
> > +	if (readdat != data) {
> > +		pr_err("Error, %s: Writing failed!: address: 0x%x, value : 0x%x,
> > +		       readval: 0x%x\n", __func__, addr, data, readdat);
> 
> 
> Doesn't need to return error value?
Yes, I think it will be changed  to return error value.
> 
> 
> > +	}
> > +}
> > +
> > +static int sdhci_cdns6_reset_phy_dll(struct sdhci_cdns_plat *plat,
> > +				     unsigned int reset)
> > +{
> > +	void __iomem *reg = plat->hrs_addr + SDHCI_CDNS_HRS09;
> > +	u32 tmp;
> > +	int ret;
> > +
> > +	tmp = readl(reg);
> > +	tmp &= ~SDHCI_CDNS_HRS09_PHY_SW_RESET;
> > +
> > +	if (reset)	/* Switch On DLL Reset */
> 
> 
> 	/* Swithc On DLL Reset */
Noted, thanks for telling.

> 	if (reset)
> 		...
> 	else
> 		...
> 
> 
> > +		tmp |= FIELD_PREP(SDHCI_CDNS_HRS09_PHY_SW_RESET, 0);
> > +	else		/* Switch Off DLL Reset */
> > +		tmp |= FIELD_PREP(SDHCI_CDNS_HRS09_PHY_SW_RESET, 1);
> > +
> > +	writel(tmp, reg);
> > +
> > +	if (!reset) {
> > +		ret = readl_poll_timeout(reg, tmp,
> > +					 (tmp &
> SDHCI_CDNS_HRS09_PHY_INIT_COMPLETE),
> > +					 3000);
> 
> Add a comment why it needs to poll during 3000.
Actually 3000 us is a safe range. User manual doesn’t specify, I put myself.
Comment will be like below:
/* After reset, wait for PHY_INIT_COMPLETE in HRS09 register until it is set to 1 within 3000us*/

> 
> > +	}
> > +
> > +	return ret;
> > +}
> > +
> > +int sdhci_cdns6_phy_init(struct sdhci_cdns_plat *plat, u32 mode) {
> > +	struct phy_reg *phy_cfgs;
> > +	struct controller_reg *ctrl_cfgs;
> > +	void __iomem *reg = NULL;
> > +	u32 tmp;
> > +	int ret;
> > +
> > +	switch (mode) {
> > +	case SDHCI_CDNS_HRS06_MODE_SD:
> > +		phy_cfgs = &sd_ds_phy_cfg;
> > +		ctrl_cfgs = &sd_ds_ctrl_cfg;
> > +		break;
> > +
> > +	case SDHCI_CDNS_HRS06_MODE_MMC_SDR:
> > +		phy_cfgs = &emmc_sdr_phy_cfg;
> > +		ctrl_cfgs = &emmc_sdr_ctrl_cfg;
> > +		break;
> > +
> > +	case SDHCI_CDNS_HRS06_MODE_MMC_DDR:
> > +		phy_cfgs = &emmc_ddr_phy_cfg;
> > +		ctrl_cfgs = &emmc_ddr_ctrl_cfg;
> > +		break;
> > +
> > +	case SDHCI_CDNS_HRS06_MODE_MMC_HS200:
> > +		phy_cfgs = &emmc_hs200_phy_cfg;
> > +		ctrl_cfgs = &emmc_hs200_ctrl_cfg;
> > +		break;
> > +
> > +	case SDHCI_CDNS_HRS06_MODE_MMC_HS400:
> > +		phy_cfgs = &emmc_hs400_phy_cfg;
> > +		ctrl_cfgs = &emmc_hs400_ctrl_cfg;
> > +		break;
> 
> 
> If there is no matched mod, phy_cfgs and ctrl_cfgs should be NULL.
Noted.

> 
> > +	}
> > +
> > +	/* Switch On the DLL Reset */
> > +	sdhci_cdns6_reset_phy_dll(plat, 1);
> > +
> > +	sdhci_cdns6_write_phy_reg(plat, PHY_DQS_TIMING_REG_ADDR,
> > +				  phy_cfgs->phy_dqs_timing);
> > +	sdhci_cdns6_write_phy_reg(plat, PHY_GATE_LPBK_CTRL_REG_ADDR,
> > +				  phy_cfgs->phy_gate_lpbk_ctrl);
> > +	sdhci_cdns6_write_phy_reg(plat, PHY_DLL_SLAVE_CTRL_REG_ADDR,
> > +				  phy_cfgs->phy_dll_slave_ctrl);
> > +
> > +	/* Switch Off the DLL Reset */
> > +	ret = sdhci_cdns6_reset_phy_dll(plat, 0);
> > +	if (ret) {
> > +		printf("sdhci_cdns6_reset_phy is not completed\n");
> > +		return ret;
> > +	}
> > +
> > +	/* Set PHY DQ TIMING control register */
> > +	sdhci_cdns6_write_phy_reg(plat, PHY_DQ_TIMING_REG_ADDR,
> > +				  phy_cfgs->phy_dq_timing);
> > +
> > +	/* Set HRS09 register */
> > +	reg = plat->hrs_addr + SDHCI_CDNS_HRS09;
> > +	tmp = readl(reg);
> 
> 
> I'm not suer why plat-hrs_addr assigned to reg.
> How about using the below?
> 
> hrs_base_addr = plat->hrs_addr;
> 
> readl(hrs_base_addr + SDHCI_CDNS_HRS10);
> 
> Then you can change the below code like
> 
> readl(hrs_base_addr + SDHCI_CDNS_HRS16); readl(hrs_base_addr +
> SDHCI_CDNS_HRS07);
> 
> Otherwise, readl(plat->hrs_addr + SDHCI_xxx);
Noted, this looked simpler.

> 
> 
> > +	tmp &= ~(SDHCI_CDNS_HRS09_EXTENDED_WR_MODE |
> > +		 SDHCI_CDNS_HRS09_EXTENDED_RD_MODE |
> > +		 SDHCI_CDNS_HRS09_RDDATA_EN |
> > +		 SDHCI_CDNS_HRS09_RDCMD_EN);
> > +	tmp |= ctrl_cfgs->hrs09;
> > +	writel(tmp, reg);
> > +
> > +	/* Set HRS10 register */
> > +	reg = plat->hrs_addr + SDHCI_CDNS_HRS10;
> > +	tmp = readl(reg);
> > +	tmp &= ~SDHCI_CDNS_HRS10_HCSDCLKADJ;
> > +	tmp |= ctrl_cfgs->hrs10;
> > +	writel(tmp, reg);
> > +
> > +	/* Set HRS16 register */
> > +	reg = plat->hrs_addr + SDHCI_CDNS_HRS16;
> > +	tmp = ctrl_cfgs->hrs16;
> > +	writel(tmp, reg);
> > +
> > +	/* Set HRS07 register */
> > +	reg = plat->hrs_addr + SDHCI_CDNS_HRS07;
> > +	tmp = ctrl_cfgs->hrs07;
> > +	writel(tmp, reg);
> > +
> > +	return 0;
> > +}
> > +
> > +int sdhci_cdns6_set_tune_val(struct sdhci_cdns_plat *plat,
> > +			     unsigned int val)
> > +{
> > +	u32 tmp, tuneval;
> 
> Ditto.
Any problem with this? I think current way looked simple.

> 
> Best Regards,
> Jaehoon Chung
> 
> > +
> > +	tuneval = (val * 256) / SDHCI_CDNS_MAX_TUNING_LOOP;
> > +
> > +	tmp = sdhci_cdns6_read_phy_reg(plat,
> PHY_DLL_SLAVE_CTRL_REG_ADDR);
> > +	tmp &= ~(PHY_DLL_SLAVE_CTRL_REG_READ_DQS_CMD_DELAY |
> > +		 PHY_DLL_SLAVE_CTRL_REG_READ_DQS_DELAY);
> > +	tmp |=
> FIELD_PREP(PHY_DLL_SLAVE_CTRL_REG_READ_DQS_CMD_DELAY, tuneval) |
> > +		FIELD_PREP(PHY_DLL_SLAVE_CTRL_REG_READ_DQS_DELAY,
> tuneval);
> > +	sdhci_cdns6_write_phy_reg(plat, PHY_DLL_SLAVE_CTRL_REG_ADDR,
> tmp);
> > +
> > +	return 0;
> > +}
> 



More information about the U-Boot mailing list