[PATCH 3/9] phy: qcom: uniphy: Update UNIPHY driver to be a common driver

Praveenkumar I quic_ipkumar at quicinc.com
Thu Aug 31 05:05:48 PDT 2023


On 8/29/2023 8:07 PM, Dmitry Baryshkov wrote:
> On Tue, 29 Aug 2023 at 17:00, Praveenkumar I <quic_ipkumar at quicinc.com> wrote:
>> This patch updates the UNIPHY driver to be a common driver to
>> accommodate all UNIPHY / Combo PHY. This driver can be used for
>> both USB and PCIe UNIPHY. Using phy-mul-sel from DTS MUX selection
>> for USB / PCIe can be acheived.
> I'm not sure why you are talking about PCIe here. This patch adds only
> SS PHY support.
>
> Also, I'd like to point out that we had this 'USB and PCIe and
> everything else' design in the QMP driver. We had to split the driver
> into individual pieces to make it manageable again.
This Uniphy 22ull is a combo PHY used between USB GEN3 or PCIe GEN3. Via 
mux selection
the one of the controller interface can use it. Hence thought to have a 
same PHY driver for
both USB3 and PCIe. Will drop the plan of adding the PCIe support.
>
>> Signed-off-by: Praveenkumar I <quic_ipkumar at quicinc.com>
>> ---
>>   drivers/phy/qualcomm/phy-qcom-uniphy.c | 401 +++++++++++++++++++++----
>>   1 file changed, 335 insertions(+), 66 deletions(-)
>>
>> diff --git a/drivers/phy/qualcomm/phy-qcom-uniphy.c b/drivers/phy/qualcomm/phy-qcom-uniphy.c
>> index da6f290af722..eb71588f5417 100644
>> --- a/drivers/phy/qualcomm/phy-qcom-uniphy.c
>> +++ b/drivers/phy/qualcomm/phy-qcom-uniphy.c
>> @@ -5,141 +5,410 @@
>>    * Based on code from
>>    * Allwinner Technology Co., Ltd. <www.allwinnertech.com>
>>    *
>> + * Modified the driver to be common for Qualcomm UNIPHYs
>> + * Copyright (c) 2023, The Linux Foundation. All rights reserved.
> I'd not call this 'modified', but rather 'rewritten from scratch.
Sure, will correct.
>
>>    */
>>
>> +#include <linux/bitfield.h>
>> +#include <linux/clk.h>
>> +#include <linux/clk-provider.h>
>>   #include <linux/delay.h>
>>   #include <linux/err.h>
>>   #include <linux/io.h>
>>   #include <linux/kernel.h>
>> +#include <linux/mfd/syscon.h>
>>   #include <linux/module.h>
>>   #include <linux/mutex.h>
>>   #include <linux/of.h>
>>   #include <linux/phy/phy.h>
>>   #include <linux/platform_device.h>
>> +#include <linux/regmap.h>
>> +#include <linux/regulator/consumer.h>
>>   #include <linux/reset.h>
>>
>> -struct ipq4019_usb_phy {
>> +struct uniphy_init_tbl {
>> +       unsigned int offset;
>> +       unsigned int val;
>> +};
> unused
>
>> +
>> +#define UNIPHY_INIT_CFG(o, v)          \
>> +       {                               \
>> +               .offset = o,            \
>> +               .val = v,               \
>> +       }
> unused
>
>> +
>> +struct uniphy_cfg {
>> +       const struct uniphy_init_tbl *init_seq;
>> +       int num_init_seq;
>> +       const char * const *clk_list;
>> +       int num_clks;
>> +       const char * const *reset_list;
>> +       int num_resets;
>> +       const char * const *vreg_list;
>> +       int num_vregs;
>> +       unsigned int pipe_clk_rate;
>> +       unsigned int reset_udelay;
>> +       unsigned int autoload_udelay;
>> +};
>> +
>> +struct qcom_uniphy {
>>          struct device           *dev;
>> +       const struct uniphy_cfg *cfg;
>>          struct phy              *phy;
>>          void __iomem            *base;
>> -       struct reset_control    *por_rst;
>> -       struct reset_control    *srif_rst;
>> +       struct clk_bulk_data    *clks;
>> +       struct reset_control_bulk_data  *resets;
>> +       struct regulator_bulk_data *vregs;
>> +       struct clk_fixed_rate pipe_clk_fixed;
>> +};
>> +
>> +static const char * const ipq4019_ssphy_reset_l[] = {
>> +       "por_rst",
>> +};
>> +
>> +static const struct uniphy_cfg ipq4019_usb_ssphy_cfg = {
>> +       .reset_list     = ipq4019_ssphy_reset_l,
>> +       .num_resets     = ARRAY_SIZE(ipq4019_ssphy_reset_l),
>> +       .reset_udelay   = 10000,
>> +
>>   };
>>
>> -static int ipq4019_ss_phy_power_off(struct phy *_phy)
>> +static const char * const ipq4019_hsphy_reset_l[] = {
>> +       "por_rst", "srif_rst",
>> +};
> +
>> +static const struct uniphy_cfg ipq4019_usb_hsphy_cfg = {
>> +       .reset_list     = ipq4019_hsphy_reset_l,
>> +       .num_resets     = ARRAY_SIZE(ipq4019_hsphy_reset_l),
>> +       .reset_udelay   = 10000,
>> +};
>> +
>> +static int phy_mux_sel(struct phy *phy)
>> +{
>> +       struct qcom_uniphy *uniphy = phy_get_drvdata(phy);
>> +       struct device *dev = uniphy->dev;
>> +       struct regmap *tcsr;
>> +       unsigned int args[2];
>> +       int ret;
>> +
>> +       tcsr = syscon_regmap_lookup_by_phandle_args(dev->of_node, "qcom,phy-mux-sel",
>> +                                                   ARRAY_SIZE(args), args);
> No. mux data should come from match_data rather than polluting DT with it.
>
>> +       if (IS_ERR(tcsr)) {
>> +               ret = PTR_ERR(tcsr);
>> +               if (ret == -ENOENT)
>> +                       return 0;
>> +
>> +               dev_err(dev, "failed to lookup syscon for phy mux %d\n", ret);
>> +               return ret;
>> +       }
>> +
>> +       /* PHY MUX registers only have this BIT0 */
> huh?
>
>> +       ret = regmap_write(tcsr, args[0], args[1]);
>> +       if (ret < 0) {
>> +               dev_err(dev, "PHY Mux selection failed: %d\n", ret);
>> +               return ret;
>> +       }
>> +
>> +       return 0;
>> +}
>> +
>> +static int uniphy_enable(struct phy *phy)
>>   {
>> -       struct ipq4019_usb_phy *phy = phy_get_drvdata(_phy);
>> +       struct qcom_uniphy *uniphy = phy_get_drvdata(phy);
>> +       const struct uniphy_cfg *cfg = uniphy->cfg;
>> +       const struct uniphy_init_tbl *tbl;
>> +       void __iomem *base = uniphy->base;
>> +       int i, ret;
>>
>> -       reset_control_assert(phy->por_rst);
>> -       msleep(10);
>> +       ret = regulator_bulk_enable(cfg->num_vregs, uniphy->vregs);
>> +       if (ret) {
>> +               dev_err(uniphy->dev, "failed to enable regulators: %d\n", ret);
>> +               return ret;
>> +       }
>> +
>> +       /* Assert all available resets */
>> +       for (i = 0; i < cfg->num_resets; i++) {
>> +               ret = reset_control_assert(uniphy->resets[i].rstc);
>> +               if (ret) {
>> +                       dev_err(uniphy->dev, "reset assert failed: %d\n", ret);
>> +                       goto err_assert_reset;
>> +               }
>> +               if (cfg->reset_udelay)
>> +                       usleep_range(cfg->reset_udelay, cfg->reset_udelay + 10);
>> +       }
> There is a whole reset_control_bulk_*() set of API. Please use it
> instead of hardcoding reset cycles.
>
>> +
>> +       /* Deassert all available resets */
>> +       for (i = 0; i < cfg->num_resets; i++) {
>> +               ret = reset_control_deassert(uniphy->resets[i].rstc);
>> +               if (ret) {
>> +                       dev_err(uniphy->dev, "reset deassert failed: %d\n", ret);
>> +                       goto err_assert_reset;
>> +               }
>> +               if (cfg->reset_udelay)
>> +                       usleep_range(cfg->reset_udelay, cfg->reset_udelay + 10);
>> +       }
>> +
>> +       ret = phy_mux_sel(phy);
>> +       if (ret < 0)
>> +               goto err_assert_reset;
>> +
>> +       ret = clk_bulk_prepare_enable(cfg->num_clks, uniphy->clks);
>> +       if (ret) {
>> +               dev_err(uniphy->dev, "failed to enable clocks: %d\n", ret);
>> +               goto err_assert_reset;
>> +       }
>> +
>> +       if (cfg->autoload_udelay)
>> +               usleep_range(cfg->autoload_udelay, cfg->autoload_udelay + 10);
>> +
>> +       if (cfg->num_init_seq) {
>> +               tbl = cfg->init_seq;
>> +               for (i = 0; i < cfg->num_init_seq; i++, tbl++)
>> +                       writel(tbl->val, base + tbl->offset);
>> +       }
> unused
>
>>          return 0;
>> +
>> +err_assert_reset:
>> +       /* Assert all available resets */
>> +       for (i = 0; i < cfg->num_resets; i++) {
>> +               reset_control_assert(uniphy->resets[i].rstc);
>> +               if (cfg->reset_udelay)
>> +                       usleep_range(cfg->reset_udelay, cfg->reset_udelay + 10);
>> +       }
>> +
>> +       return ret;
>>   }
>>
>> -static int ipq4019_ss_phy_power_on(struct phy *_phy)
>> +static int uniphy_disable(struct phy *phy)
>>   {
>> -       struct ipq4019_usb_phy *phy = phy_get_drvdata(_phy);
>> +       struct qcom_uniphy *uniphy = phy_get_drvdata(phy);
>> +       const struct uniphy_cfg *cfg = uniphy->cfg;
>> +       int i;
>>
>> -       ipq4019_ss_phy_power_off(_phy);
>> +       /* Assert all available resets */
>> +       for (i = 0; i < cfg->num_resets; i++) {
>> +               reset_control_assert(uniphy->resets[i].rstc);
>> +               if (cfg->reset_udelay)
>> +                       usleep_range(cfg->reset_udelay, cfg->reset_udelay + 10);
>> +       }
>>
>> -       reset_control_deassert(phy->por_rst);
>> +       clk_bulk_disable_unprepare(cfg->num_clks, uniphy->clks);
>> +
>> +       regulator_bulk_disable(cfg->num_vregs, uniphy->vregs);
>>
>>          return 0;
>>   }
>>
>> -static const struct phy_ops ipq4019_usb_ss_phy_ops = {
>> -       .power_on       = ipq4019_ss_phy_power_on,
>> -       .power_off      = ipq4019_ss_phy_power_off,
>> +static const struct phy_ops uniphy_phy_ops = {
>> +       .power_on       = uniphy_enable,
>> +       .power_off      = uniphy_disable,
> Using _enable / _disable for power_on() and power_off() isn't logical.
>
>> +       .owner          = THIS_MODULE,
>>   };
>>
>> -static int ipq4019_hs_phy_power_off(struct phy *_phy)
>> +static int qcom_uniphy_vreg_init(struct qcom_uniphy *uniphy)
>> +{
>> +       const struct uniphy_cfg *cfg = uniphy->cfg;
>> +       struct device *dev = uniphy->dev;
>> +       int i, ret;
>> +
>> +       uniphy->vregs = devm_kcalloc(dev, cfg->num_vregs,
>> +                                    sizeof(*uniphy->vregs), GFP_KERNEL);
> You know the maximum amount of regulators. Can you use an array
> instead of allocating data?
>
>> +       if (!uniphy->vregs)
>> +               return -ENOMEM;
>> +
>> +       for (i = 0; i < cfg->num_vregs; i++)
>> +               uniphy->vregs[i].supply = cfg->vreg_list[i];
>> +
>> +       ret = devm_regulator_bulk_get(dev, cfg->num_vregs, uniphy->vregs);
>> +
> Drop empty lines between ret assignment and check.
>
>> +       if (ret)
>> +               return dev_err_probe(dev, ret, "failed to get regulators\n");
>> +
>> +       return 0;
>> +}
>> +
>> +static int qcom_uniphy_reset_init(struct qcom_uniphy *uniphy)
>>   {
>> -       struct ipq4019_usb_phy *phy = phy_get_drvdata(_phy);
>> +       const struct uniphy_cfg *cfg = uniphy->cfg;
>> +       struct device *dev = uniphy->dev;
>> +       int i, ret;
>> +
>> +       uniphy->resets = devm_kcalloc(dev, cfg->num_resets,
>> +                                     sizeof(*uniphy->resets), GFP_KERNEL);
> Same here, can you use an array?
>
>> +       if (!uniphy->resets)
>> +               return -ENOMEM;
>>
>> -       reset_control_assert(phy->por_rst);
>> -       msleep(10);
>> +       for (i = 0; i < cfg->num_resets; i++)
>> +               uniphy->resets[i].id = cfg->reset_list[i];
> Declare common resets list and use
> devm_reset_control_bulk_get_optional_exclusive().
>
>> -       reset_control_assert(phy->srif_rst);
>> -       msleep(10);
>> +       ret = devm_reset_control_bulk_get_exclusive(dev, cfg->num_resets, uniphy->resets);
>> +       if (ret)
>> +               return dev_err_probe(dev, ret, "failed to get resets\n");
>>
>>          return 0;
>>   }
>>
>> -static int ipq4019_hs_phy_power_on(struct phy *_phy)
>> +static int qcom_uniphy_clk_init(struct qcom_uniphy *uniphy)
> I don't see clocks actually being used. Please do not introduce unused features.
>
>>   {
>> -       struct ipq4019_usb_phy *phy = phy_get_drvdata(_phy);
>> +       const struct uniphy_cfg *cfg = uniphy->cfg;
>> +       struct device *dev = uniphy->dev;
>> +       int i, ret;
>>
>> -       ipq4019_hs_phy_power_off(_phy);
>>
>> -       reset_control_deassert(phy->srif_rst);
>> -       msleep(10);
>> +       uniphy->clks = devm_kcalloc(dev, cfg->num_clks,
>> +                                   sizeof(*uniphy->clks), GFP_KERNEL);
>> +       if (!uniphy->clks)
>> +               return -ENOMEM;
>> +
>> +       for (i = 0; i < cfg->num_clks; i++)
>> +               uniphy->clks[i].id = cfg->clk_list[i];
>>
>> -       reset_control_deassert(phy->por_rst);
>> +       ret = devm_clk_bulk_get(dev, cfg->num_clks, uniphy->clks);
>> +       if (ret)
>> +               return dev_err_probe(dev, ret, "failed to get clocks\n");
>>
>>          return 0;
>>   }
>>
>> -static const struct phy_ops ipq4019_usb_hs_phy_ops = {
>> -       .power_on       = ipq4019_hs_phy_power_on,
>> -       .power_off      = ipq4019_hs_phy_power_off,
>> -};
>> +static void phy_clk_release_provider(void *res)
>> +{
>> +       of_clk_del_provider(res);
>> +}
>>
>> -static const struct of_device_id ipq4019_usb_phy_of_match[] = {
>> -       { .compatible = "qcom,usb-hs-ipq4019-phy", .data = &ipq4019_usb_hs_phy_ops},
>> -       { .compatible = "qcom,usb-ss-ipq4019-phy", .data = &ipq4019_usb_ss_phy_ops},
>> -       { },
>> -};
>> -MODULE_DEVICE_TABLE(of, ipq4019_usb_phy_of_match);
>> +/*
>> + * Register a fixed rate pipe clock.
>> + *
>> + * The <s>_pipe_clksrc generated by PHY goes to the GCC that gate
>> + * controls it. The <s>_pipe_clk coming out of the GCC is requested
>> + * by the PHY driver for its operations.
>> + * We register the <s>_pipe_clksrc here. The gcc driver takes care
>> + * of assigning this <s>_pipe_clksrc as parent to <s>_pipe_clk.
>> + * Below picture shows this relationship.
>> + *
>> + *         +---------------+
>> + *         |   PHY block   |<<---------------------------------------+
>> + *         |               |                                         |
>> + *         |   +-------+   |                   +-----+               |
>> + *   I/P---^-->|  PLL  |---^--->pipe_clksrc--->| GCC |--->pipe_clk---+
>> + *    clk  |   +-------+   |                   +-----+
>> + *         +---------------+
>> + */
>> +static int phy_pipe_clk_register(struct qcom_uniphy *uniphy, struct device_node *np)
>> +{
>> +       struct clk_fixed_rate *fixed = &uniphy->pipe_clk_fixed;
>> +       const struct uniphy_cfg *cfg = uniphy->cfg;
>> +       struct device *dev = uniphy->dev;
>> +       struct clk_init_data init = { };
>> +       int ret;
>> +
>> +       ret = of_property_read_string(np, "clock-output-names", &init.name);
>> +       if (ret) {
>> +               dev_err(dev, "%pOFn: No clock-output-names\n", np);
>> +               return ret;
>> +       }
>> +
>> +       init.ops = &clk_fixed_rate_ops;
>> +
>> +       fixed->fixed_rate = cfg->pipe_clk_rate;
>> +       fixed->hw.init = &init;
>>
>> -static int ipq4019_usb_phy_probe(struct platform_device *pdev)
>> +       ret = devm_clk_hw_register(dev, &fixed->hw);
>> +       if (ret)
>> +               return ret;
>> +
>> +       ret = of_clk_add_hw_provider(np, of_clk_hw_simple_get, &fixed->hw);
>> +       if (ret)
>> +               return ret;
> When you c&p something, please take care to understand the code you are copying.
> Unlike QMP drivers you can (and should) use devm_of_clk_add_hw_provider() here.
>
> Not to mention that pipe clocks are not in this patch.
>
>> +
>> +       /*
>> +        * Roll a devm action because the clock provider is the child node, but
>> +        * the child node is not actually a device.
>> +        */
>> +       return devm_add_action_or_reset(dev, phy_clk_release_provider, np);
>> +}
>> +
>> +static int qcom_uniphy_probe(struct platform_device *pdev)
>>   {
>>          struct device *dev = &pdev->dev;
>>          struct phy_provider *phy_provider;
>> -       struct ipq4019_usb_phy *phy;
>> +       struct qcom_uniphy *uniphy;
>> +       struct device_node *np;
>> +       int ret;
>>
>> -       phy = devm_kzalloc(dev, sizeof(*phy), GFP_KERNEL);
>> -       if (!phy)
>> +       uniphy = devm_kzalloc(dev, sizeof(*uniphy), GFP_KERNEL);
>> +       if (!uniphy)
>>                  return -ENOMEM;
>>
>> -       phy->dev = &pdev->dev;
>> -       phy->base = devm_platform_ioremap_resource(pdev, 0);
>> -       if (IS_ERR(phy->base)) {
>> -               dev_err(dev, "failed to remap register memory\n");
>> -               return PTR_ERR(phy->base);
>> -       }
>> +       uniphy->dev = dev;
>>
>> -       phy->por_rst = devm_reset_control_get(phy->dev, "por_rst");
>> -       if (IS_ERR(phy->por_rst)) {
>> -               if (PTR_ERR(phy->por_rst) != -EPROBE_DEFER)
>> -                       dev_err(dev, "POR reset is missing\n");
>> -               return PTR_ERR(phy->por_rst);
>> +       uniphy->cfg = of_device_get_match_data(dev);
>> +       if (!uniphy->cfg)
>> +               return -EINVAL;
>> +
>> +       uniphy->base = devm_platform_ioremap_resource(pdev, 0);
>> +       if (IS_ERR(uniphy->base)) {
>> +               ret = PTR_ERR(uniphy->base);
>> +               dev_err_probe(dev, ret, "failed to remap register memory\n");
>> +               return ret;
>>          }
>>
>> -       phy->srif_rst = devm_reset_control_get_optional(phy->dev, "srif_rst");
>> -       if (IS_ERR(phy->srif_rst))
>> -               return PTR_ERR(phy->srif_rst);
>> +       ret = qcom_uniphy_clk_init(uniphy);
>> +       if (ret)
>> +               return ret;
>> +
>> +       ret = qcom_uniphy_reset_init(uniphy);
>> +       if (ret)
>> +               return ret;
>> +
>> +       ret = qcom_uniphy_vreg_init(uniphy);
>> +       if (ret < 0)
>> +               return ret;
>> +
>> +       if (uniphy->cfg->pipe_clk_rate) {
>> +               np = of_node_get(dev->of_node);
> What for? Do you think that the driver can outlive struct device?
>
>> +               ret = phy_pipe_clk_register(uniphy, np);
>> +               if (ret) {
>> +                       dev_err_probe(dev, ret, "failed to register pipe clk\n");
>> +                       goto err;
>> +               }
>> +       }
>>
>> -       phy->phy = devm_phy_create(dev, NULL, of_device_get_match_data(dev));
>> -       if (IS_ERR(phy->phy)) {
>> -               dev_err(dev, "failed to create PHY\n");
>> -               return PTR_ERR(phy->phy);
>> +       uniphy->phy = devm_phy_create(dev, NULL, &uniphy_phy_ops);
>> +       if (IS_ERR(uniphy->phy)) {
>> +               ret = PTR_ERR(uniphy->phy);
>> +               dev_err_probe(dev, ret, "failed to create PHY\n");
>> +               goto err;
>>          }
>> -       phy_set_drvdata(phy->phy, phy);
>> +
>> +       phy_set_drvdata(uniphy->phy, uniphy);
>>
>>          phy_provider = devm_of_phy_provider_register(dev, of_phy_simple_xlate);
>>
>> -       return PTR_ERR_OR_ZERO(phy_provider);
>> +       ret = PTR_ERR_OR_ZERO(phy_provider);
>> +
>> +err:
>> +       if (uniphy->cfg->pipe_clk_rate)
>> +               of_node_put(np);
>> +       return ret;
>>   }
>>
>> -static struct platform_driver ipq4019_usb_phy_driver = {
>> -       .probe  = ipq4019_usb_phy_probe,
>> +static const struct of_device_id qcom_uniphy_of_match[] = {
>> +       { .compatible = "qcom,usb-hs-ipq4019-phy", .data = &ipq4019_usb_hsphy_cfg},
>> +       { .compatible = "qcom,usb-ss-ipq4019-phy", .data = &ipq4019_usb_ssphy_cfg},
>> +       { },
>> +};
>> +MODULE_DEVICE_TABLE(of, qcom_uniphy_of_match);
>> +
>> +static struct platform_driver qcom_uniphy_driver = {
>> +       .probe  = qcom_uniphy_probe,
>>          .driver = {
>> -               .of_match_table = ipq4019_usb_phy_of_match,
>> -               .name  = "ipq4019-usb-phy",
>> +               .of_match_table = qcom_uniphy_of_match,
>> +               .name  = "qcom-uniphy",
>>          }
>>   };
>> -module_platform_driver(ipq4019_usb_phy_driver);
>> +module_platform_driver(qcom_uniphy_driver);
>>
>> -MODULE_DESCRIPTION("QCOM/IPQ4019 USB phy driver");
>> +MODULE_DESCRIPTION("QCOM uniphy driver");
>>   MODULE_AUTHOR("John Crispin <john at phrozen.org>");
>>   MODULE_LICENSE("GPL v2");
> General comment: please consider dropping this beast and starting from
> scratch, adding only really necessary bits to the existing ipq4019 USB
> PHY driver.
Rewritten the entire driver considering the IPQ5332 as well and because 
of it
there are unused code which are utilized in patch [8/9]. I should have 
added only the IPQ4019
support alone in this patch and then incrementally update it for 
IPQ5332. Will follow that here
onwards. Will address all the review comments.

- Praveenkumar



More information about the linux-arm-kernel mailing list