phy: intel: Add Keem Bay eMMC PHY support
authorWan Ahmad Zainie <wan.ahmad.zainie.wan.mohamad@intel.com>
Sun, 13 Sep 2020 23:55:22 +0000 (07:55 +0800)
committerVinod Koul <vkoul@kernel.org>
Wed, 16 Sep 2020 12:15:19 +0000 (17:45 +0530)
Add support for eMMC PHY on Intel Keem Bay SoC.

Signed-off-by: Wan Ahmad Zainie <wan.ahmad.zainie.wan.mohamad@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Link: https://lore.kernel.org/r/20200913235522.4316-4-wan.ahmad.zainie.wan.mohamad@intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
drivers/phy/intel/Kconfig
drivers/phy/intel/Makefile
drivers/phy/intel/phy-intel-keembay-emmc.c [new file with mode: 0644]

index db8586c..58ec695 100644 (file)
@@ -2,6 +2,18 @@
 #
 # Phy drivers for Intel platforms
 #
+config PHY_INTEL_KEEMBAY_EMMC
+       tristate "Intel Keem Bay EMMC PHY driver"
+       depends on (OF && ARM64) || COMPILE_TEST
+       depends on HAS_IOMEM
+       select GENERIC_PHY
+       select REGMAP_MMIO
+       help
+         Choose this option if you have an Intel Keem Bay SoC.
+
+         To compile this driver as a module, choose M here: the module
+         will be called phy-keembay-emmc.ko.
+
 config PHY_INTEL_LGM_COMBO
        bool "Intel Lightning Mountain ComboPHY driver"
        depends on X86 || COMPILE_TEST
index 662385d..a5e0af5 100644 (file)
@@ -1,3 +1,4 @@
 # SPDX-License-Identifier: GPL-2.0
+obj-$(CONFIG_PHY_INTEL_KEEMBAY_EMMC)   += phy-intel-keembay-emmc.o
 obj-$(CONFIG_PHY_INTEL_LGM_COMBO)      += phy-intel-lgm-combo.o
 obj-$(CONFIG_PHY_INTEL_LGM_EMMC)       += phy-intel-lgm-emmc.o
diff --git a/drivers/phy/intel/phy-intel-keembay-emmc.c b/drivers/phy/intel/phy-intel-keembay-emmc.c
new file mode 100644 (file)
index 0000000..eb7c635
--- /dev/null
@@ -0,0 +1,307 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Intel Keem Bay eMMC PHY driver
+ * Copyright (C) 2020 Intel Corporation
+ */
+
+#include <linux/bitfield.h>
+#include <linux/clk.h>
+#include <linux/delay.h>
+#include <linux/mfd/syscon.h>
+#include <linux/module.h>
+#include <linux/of.h>
+#include <linux/of_address.h>
+#include <linux/phy/phy.h>
+#include <linux/platform_device.h>
+#include <linux/regmap.h>
+
+/* eMMC/SD/SDIO core/phy configuration registers */
+#define PHY_CFG_0              0x24
+#define  SEL_DLY_TXCLK_MASK    BIT(29)
+#define  OTAP_DLY_ENA_MASK     BIT(27)
+#define  OTAP_DLY_SEL_MASK     GENMASK(26, 23)
+#define  DLL_EN_MASK           BIT(10)
+#define  PWR_DOWN_MASK         BIT(0)
+
+#define PHY_CFG_2              0x2c
+#define  SEL_FREQ_MASK         GENMASK(12, 10)
+
+#define PHY_STAT               0x40
+#define  CAL_DONE_MASK         BIT(6)
+#define  IS_CALDONE(x)         ((x) & CAL_DONE_MASK)
+#define  DLL_RDY_MASK          BIT(5)
+#define  IS_DLLRDY(x)          ((x) & DLL_RDY_MASK)
+
+/* From ACS_eMMC51_16nFFC_RO1100_Userguide_v1p0.pdf p17 */
+#define FREQSEL_200M_170M      0x0
+#define FREQSEL_170M_140M      0x1
+#define FREQSEL_140M_110M      0x2
+#define FREQSEL_110M_80M       0x3
+#define FREQSEL_80M_50M                0x4
+
+struct keembay_emmc_phy {
+       struct regmap *syscfg;
+       struct clk *emmcclk;
+};
+
+static const struct regmap_config keembay_regmap_config = {
+       .reg_bits = 32,
+       .val_bits = 32,
+       .reg_stride = 4,
+};
+
+static int keembay_emmc_phy_power(struct phy *phy, bool on_off)
+{
+       struct keembay_emmc_phy *priv = phy_get_drvdata(phy);
+       unsigned int caldone;
+       unsigned int dllrdy;
+       unsigned int freqsel;
+       unsigned int mhz;
+       int ret;
+
+       /*
+        * Keep phyctrl_pdb and phyctrl_endll low to allow
+        * initialization of CALIO state M/C DFFs
+        */
+       ret = regmap_update_bits(priv->syscfg, PHY_CFG_0, PWR_DOWN_MASK,
+                                FIELD_PREP(PWR_DOWN_MASK, 0));
+       if (ret) {
+               dev_err(&phy->dev, "CALIO power down bar failed: %d\n", ret);
+               return ret;
+       }
+
+       ret = regmap_update_bits(priv->syscfg, PHY_CFG_0, DLL_EN_MASK,
+                                FIELD_PREP(DLL_EN_MASK, 0));
+       if (ret) {
+               dev_err(&phy->dev, "turn off the dll failed: %d\n", ret);
+               return ret;
+       }
+
+       /* Already finish power off above */
+       if (!on_off)
+               return 0;
+
+       mhz = DIV_ROUND_CLOSEST(clk_get_rate(priv->emmcclk), 1000000);
+       if (mhz <= 200 && mhz >= 170)
+               freqsel = FREQSEL_200M_170M;
+       else if (mhz <= 170 && mhz >= 140)
+               freqsel = FREQSEL_170M_140M;
+       else if (mhz <= 140 && mhz >= 110)
+               freqsel = FREQSEL_140M_110M;
+       else if (mhz <= 110 && mhz >= 80)
+               freqsel = FREQSEL_110M_80M;
+       else if (mhz <= 80 && mhz >= 50)
+               freqsel = FREQSEL_80M_50M;
+       else
+               freqsel = 0x0;
+
+       if (mhz < 50 || mhz > 200)
+               dev_warn(&phy->dev, "Unsupported rate: %d MHz\n", mhz);
+
+       /*
+        * According to the user manual, calpad calibration
+        * cycle takes more than 2us without the minimal recommended
+        * value, so we may need a little margin here
+        */
+       udelay(5);
+
+       ret = regmap_update_bits(priv->syscfg, PHY_CFG_0, PWR_DOWN_MASK,
+                                FIELD_PREP(PWR_DOWN_MASK, 1));
+       if (ret) {
+               dev_err(&phy->dev, "CALIO power down bar failed: %d\n", ret);
+               return ret;
+       }
+
+       /*
+        * According to the user manual, it asks driver to wait 5us for
+        * calpad busy trimming. However it is documented that this value is
+        * PVT(A.K.A. process, voltage and temperature) relevant, so some
+        * failure cases are found which indicates we should be more tolerant
+        * to calpad busy trimming.
+        */
+       ret = regmap_read_poll_timeout(priv->syscfg, PHY_STAT,
+                                      caldone, IS_CALDONE(caldone),
+                                      0, 50);
+       if (ret) {
+               dev_err(&phy->dev, "caldone failed, ret=%d\n", ret);
+               return ret;
+       }
+
+       /* Set the frequency of the DLL operation */
+       ret = regmap_update_bits(priv->syscfg, PHY_CFG_2, SEL_FREQ_MASK,
+                                FIELD_PREP(SEL_FREQ_MASK, freqsel));
+       if (ret) {
+               dev_err(&phy->dev, "set the frequency of dll failed:%d\n", ret);
+               return ret;
+       }
+
+       /* Turn on the DLL */
+       ret = regmap_update_bits(priv->syscfg, PHY_CFG_0, DLL_EN_MASK,
+                                FIELD_PREP(DLL_EN_MASK, 1));
+       if (ret) {
+               dev_err(&phy->dev, "turn on the dll failed: %d\n", ret);
+               return ret;
+       }
+
+       /*
+        * We turned on the DLL even though the rate was 0 because we the
+        * clock might be turned on later.  ...but we can't wait for the DLL
+        * to lock when the rate is 0 because it will never lock with no
+        * input clock.
+        *
+        * Technically we should be checking the lock later when the clock
+        * is turned on, but for now we won't.
+        */
+       if (mhz == 0)
+               return 0;
+
+       /*
+        * After enabling analog DLL circuits docs say that we need 10.2 us if
+        * our source clock is at 50 MHz and that lock time scales linearly
+        * with clock speed. If we are powering on the PHY and the card clock
+        * is super slow (like 100kHz) this could take as long as 5.1 ms as
+        * per the math: 10.2 us * (50000000 Hz / 100000 Hz) => 5.1 ms
+        * hopefully we won't be running at 100 kHz, but we should still make
+        * sure we wait long enough.
+        *
+        * NOTE: There appear to be corner cases where the DLL seems to take
+        * extra long to lock for reasons that aren't understood. In some
+        * extreme cases we've seen it take up to over 10ms (!). We'll be
+        * generous and give it 50ms.
+        */
+       ret = regmap_read_poll_timeout(priv->syscfg, PHY_STAT,
+                                      dllrdy, IS_DLLRDY(dllrdy),
+                                      0, 50 * USEC_PER_MSEC);
+       if (ret)
+               dev_err(&phy->dev, "dllrdy failed, ret=%d\n", ret);
+
+       return ret;
+}
+
+static int keembay_emmc_phy_init(struct phy *phy)
+{
+       struct keembay_emmc_phy *priv = phy_get_drvdata(phy);
+
+       /*
+        * We purposely get the clock here and not in probe to avoid the
+        * circular dependency problem. We expect:
+        * - PHY driver to probe
+        * - SDHCI driver to start probe
+        * - SDHCI driver to register it's clock
+        * - SDHCI driver to get the PHY
+        * - SDHCI driver to init the PHY
+        *
+        * The clock is optional, so upon any error just return it like
+        * any other error to user.
+        */
+       priv->emmcclk = clk_get_optional(&phy->dev, "emmcclk");
+
+       return PTR_ERR_OR_ZERO(priv->emmcclk);
+}
+
+static int keembay_emmc_phy_exit(struct phy *phy)
+{
+       struct keembay_emmc_phy *priv = phy_get_drvdata(phy);
+
+       clk_put(priv->emmcclk);
+
+       return 0;
+};
+
+static int keembay_emmc_phy_power_on(struct phy *phy)
+{
+       struct keembay_emmc_phy *priv = phy_get_drvdata(phy);
+       int ret;
+
+       /* Delay chain based txclk: enable */
+       ret = regmap_update_bits(priv->syscfg, PHY_CFG_0, SEL_DLY_TXCLK_MASK,
+                                FIELD_PREP(SEL_DLY_TXCLK_MASK, 1));
+       if (ret) {
+               dev_err(&phy->dev, "ERROR: delay chain txclk set: %d\n", ret);
+               return ret;
+       }
+
+       /* Output tap delay: enable */
+       ret = regmap_update_bits(priv->syscfg, PHY_CFG_0, OTAP_DLY_ENA_MASK,
+                                FIELD_PREP(OTAP_DLY_ENA_MASK, 1));
+       if (ret) {
+               dev_err(&phy->dev, "ERROR: output tap delay set: %d\n", ret);
+               return ret;
+       }
+
+       /* Output tap delay */
+       ret = regmap_update_bits(priv->syscfg, PHY_CFG_0, OTAP_DLY_SEL_MASK,
+                                FIELD_PREP(OTAP_DLY_SEL_MASK, 2));
+       if (ret) {
+               dev_err(&phy->dev, "ERROR: output tap delay select: %d\n", ret);
+               return ret;
+       }
+
+       /* Power up eMMC phy analog blocks */
+       return keembay_emmc_phy_power(phy, true);
+}
+
+static int keembay_emmc_phy_power_off(struct phy *phy)
+{
+       /* Power down eMMC phy analog blocks */
+       return keembay_emmc_phy_power(phy, false);
+}
+
+static const struct phy_ops ops = {
+       .init           = keembay_emmc_phy_init,
+       .exit           = keembay_emmc_phy_exit,
+       .power_on       = keembay_emmc_phy_power_on,
+       .power_off      = keembay_emmc_phy_power_off,
+       .owner          = THIS_MODULE,
+};
+
+static int keembay_emmc_phy_probe(struct platform_device *pdev)
+{
+       struct device *dev = &pdev->dev;
+       struct device_node *np = dev->of_node;
+       struct keembay_emmc_phy *priv;
+       struct phy *generic_phy;
+       struct phy_provider *phy_provider;
+       void __iomem *base;
+
+       priv = devm_kzalloc(dev, sizeof(*priv), GFP_KERNEL);
+       if (!priv)
+               return -ENOMEM;
+
+       base = devm_platform_ioremap_resource(pdev, 0);
+       if (IS_ERR(base))
+               return PTR_ERR(base);
+
+       priv->syscfg = devm_regmap_init_mmio(dev, base, &keembay_regmap_config);
+       if (IS_ERR(priv->syscfg))
+               return PTR_ERR(priv->syscfg);
+
+       generic_phy = devm_phy_create(dev, np, &ops);
+       if (IS_ERR(generic_phy))
+               return dev_err_probe(dev, PTR_ERR(generic_phy),
+                                    "failed to create PHY\n");
+
+       phy_set_drvdata(generic_phy, priv);
+       phy_provider = devm_of_phy_provider_register(dev, of_phy_simple_xlate);
+
+       return PTR_ERR_OR_ZERO(phy_provider);
+}
+
+static const struct of_device_id keembay_emmc_phy_dt_ids[] = {
+       { .compatible = "intel,keembay-emmc-phy" },
+       {}
+};
+MODULE_DEVICE_TABLE(of, keembay_emmc_phy_dt_ids);
+
+static struct platform_driver keembay_emmc_phy_driver = {
+       .probe          = keembay_emmc_phy_probe,
+       .driver         = {
+               .name   = "keembay-emmc-phy",
+               .of_match_table = keembay_emmc_phy_dt_ids,
+       },
+};
+module_platform_driver(keembay_emmc_phy_driver);
+
+MODULE_AUTHOR("Wan Ahmad Zainie <wan.ahmad.zainie.wan.mohamad@intel.com>");
+MODULE_DESCRIPTION("Intel Keem Bay eMMC PHY driver");
+MODULE_LICENSE("GPL v2");