Re: [PATCH v5 3/3] mmc: host: meson-mx-sdhc: new driver for the Amlogic Meson SDHC host

[Ulf Hansson <ulf.hansson@linaro.org>]

On Sat, 28 Mar 2020 at 01:33, Martin Blumenstingl
<martin.blumenstingl@googlemail.com> wrote:
>
> The SDHC MMC host controller on Amlogic SoCs provides an eMMC and MMC
> card interface with 1/4/8-bit bus width.
> It supports eMMC spec 4.4x/4.5x including HS200 (up to 100MHz clock).
>
> The public S805 datasheet [0] contains a short documentation about the
> registers. Unfortunately it does not describe how to use the registers
> to make the hardware work. Thus this driver is based on reading (and
> understanding) the Amlogic 3.10 GPL kernel code.
>
> Some hardware details are not easy to see. Jianxin Pan was kind enough
> to answer my questions:
> The hardware has built-in busy timeout support. The maximum timeout is
> 30 seconds. This is only documented in Amlogic's internal
> documentation.
>
> The controller only works with very specific clock configurations. The
> details are not part of the public datasheet. In my own words the
> supported configurations are:
> - 399.812kHz:   clkin =  850MHz div = 2126 sd_rx_phase = 63
> - 1MHz:         clkin =  850MHz div = 850  sd_rx_phase = 55
> - 5.986MHz:     clkin =  850MHz div = 142  sd_rx_phase = 24
> - 25MHz:        clkin =  850MHz div = 34   sd_rx_phase = 15
> - 47.222MHz:    clkin =  850MHz div = 18   sd_rx_phase = 11/15 (SDR50/HS)
> - 53.125MHz:    clkin =  850MHz div = 16   sd_rx_phase = (tuning)
> - 70.833MHz:    clkin =  850MHz div = 12   sd_rx_phase = (tuning)
> - 85MHz:        clkin =  850MHz div = 10   sd_rx_phase = (tuning)
> - 94.44MHz:     clkin =  850MHz div = 9    sd_rx_phase = (tuning)
> - 106.25MHz:    clkin =  850MHz div = 8    sd_rx_phase = (tuning)
> - 127.5MHz:     clkin = 1275MHz div = 10   sd_rx_phase = (tuning)
> - 141.667MHz:   clkin =  850MHz div = 6    sd_rx_phase = (tuning)
> - 159.375MHz:   clkin = 1275MHz div = 8    sd_rx_phase = (tuning)
> - 212.5MHz:     clkin = 1275MHz div = 6    sd_rx_phase = (tuning)
> - (sd_tx_phase is always 1, 94.44MHz is not listed in the datasheet
>    but this is what the 3.10 BSP kernel on Odroid-C1 actually uses)
>
> NOTE: CMD23 support is disabled for now because it results in command
> timeouts and thus decreases read performance.
>
> Tested-by: Wei Wang <lnykww@gmail.com>
> Tested-by: Xin Yin <yinxin_1989@aliyun.com>
> Reviewed-by: Xin Yin <yinxin_1989@aliyun.com>
> Signed-off-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
> ---
>  drivers/mmc/host/Kconfig         |   14 +
>  drivers/mmc/host/Makefile        |    1 +
>  drivers/mmc/host/meson-mx-sdhc.c | 1064 ++++++++++++++++++++++++++++++
>  3 files changed, 1079 insertions(+)
>  create mode 100644 drivers/mmc/host/meson-mx-sdhc.c
>
> diff --git a/drivers/mmc/host/Kconfig b/drivers/mmc/host/Kconfig
> index 3a5089f0332c..2c8361aa7eed 100644
> --- a/drivers/mmc/host/Kconfig
> +++ b/drivers/mmc/host/Kconfig
> @@ -405,6 +405,20 @@ config MMC_MESON_GX
>
>           If you have a controller with this interface, say Y here.
>
> +config MMC_MESON_MX_SDHC
> +       tristate "Amlogic Meson SDHC Host Controller support"
> +       depends on (ARM && ARCH_MESON) || COMPILE_TEST
> +       depends on OF
> +       imply COMMON_CLK_MESON_MX_SDHC
> +       help
> +         This selects support for the SDHC Host Controller on
> +         Amlogic Meson6, Meson8, Meson8b and Meson8m2 SoCs.
> +         The controller supports the SD/SDIO Spec 3.x and eMMC Spec 4.5x
> +         with 1, 4, and 8 bit bus widths.
> +
> +         If you have a controller with this interface, say Y or M here.
> +         If unsure, say N.
> +
>  config MMC_MESON_MX_SDIO
>         tristate "Amlogic Meson6/Meson8/Meson8b SD/MMC Host Controller support"
>         depends on ARCH_MESON || COMPILE_TEST
> diff --git a/drivers/mmc/host/Makefile b/drivers/mmc/host/Makefile
> index 21d9089e5eda..9a0c22a8139a 100644
> --- a/drivers/mmc/host/Makefile
> +++ b/drivers/mmc/host/Makefile
> @@ -68,6 +68,7 @@ obj-$(CONFIG_MMC_VUB300)      += vub300.o
>  obj-$(CONFIG_MMC_USHC)         += ushc.o
>  obj-$(CONFIG_MMC_WMT)          += wmt-sdmmc.o
>  obj-$(CONFIG_MMC_MESON_GX)     += meson-gx-mmc.o
> +obj-$(CONFIG_MMC_MESON_MX_SDHC)        += meson-mx-sdhc.o
>  obj-$(CONFIG_MMC_MESON_MX_SDIO)        += meson-mx-sdio.o
>  obj-$(CONFIG_MMC_MOXART)       += moxart-mmc.o
>  obj-$(CONFIG_MMC_SUNXI)                += sunxi-mmc.o
> diff --git a/drivers/mmc/host/meson-mx-sdhc.c b/drivers/mmc/host/meson-mx-sdhc.c
> new file mode 100644
> index 000000000000..b92400ee488b
> --- /dev/null
> +++ b/drivers/mmc/host/meson-mx-sdhc.c

[...]

> +
> +static void meson_mx_sdhc_wait_cmd_ready(struct mmc_host *mmc)
> +{
> +       struct meson_mx_sdhc_host *host = mmc_priv(mmc);
> +       u32 stat, esta;
> +       int ret;
> +
> +       ret = regmap_read_poll_timeout(host->regmap, MESON_SDHC_STAT, stat,
> +                                      !(stat & MESON_SDHC_STAT_CMD_BUSY), 1,
> +                                      100000);

Please use defines for timeout values.

> +       if (ret) {
> +               dev_warn(mmc_dev(mmc),
> +                        "Failed to poll for CMD_BUSY while processing CMD%d\n",
> +                        host->cmd->opcode);
> +               meson_mx_sdhc_hw_reset(mmc);
> +       }
> +
> +       ret = regmap_read_poll_timeout(host->regmap, MESON_SDHC_ESTA, esta,
> +                                      !(esta & MESON_SDHC_ESTA_11_13), 1,
> +                                      100000);

Ditto.

> +       if (ret) {
> +               dev_warn(mmc_dev(mmc),
> +                        "Failed to poll for ESTA[13:11] while processing CMD%d\n",
> +                        host->cmd->opcode);
> +               meson_mx_sdhc_hw_reset(mmc);
> +       }
> +}
> +
> +static void meson_mx_sdhc_start_cmd(struct mmc_host *mmc,
> +                                   struct mmc_command *cmd)
> +{
> +       struct meson_mx_sdhc_host *host = mmc_priv(mmc);
> +       u32 ictl, send;
> +       int pack_len;
> +
> +       host->cmd = cmd;
> +
> +       ictl = MESON_SDHC_ICTL_DATA_TIMEOUT | MESON_SDHC_ICTL_DATA_ERR_CRC |
> +              MESON_SDHC_ICTL_RXFIFO_FULL | MESON_SDHC_ICTL_TXFIFO_EMPTY |
> +              MESON_SDHC_ICTL_RESP_TIMEOUT | MESON_SDHC_ICTL_RESP_ERR_CRC;
> +
> +       send = FIELD_PREP(MESON_SDHC_SEND_CMD_INDEX, cmd->opcode);
> +
> +       if (cmd->data) {
> +               send |= MESON_SDHC_SEND_CMD_HAS_DATA;
> +               send |= FIELD_PREP(MESON_SDHC_SEND_TOTAL_PACK,
> +                                  cmd->data->blocks - 1);
> +
> +               if (cmd->data->blksz < MESON_SDHC_MAX_BLK_SIZE)
> +                       pack_len = cmd->data->blksz;
> +               else
> +                       pack_len = 0;
> +
> +               if (cmd->data->flags & MMC_DATA_WRITE)
> +                       send |= MESON_SDHC_SEND_DATA_DIR;
> +
> +               /*
> +                * If command with no data, just wait response done
> +                * interrupt(int[0]), and if command with data transfer, just
> +                * wait dma done interrupt(int[11]), don't need care about
> +                * dat0 busy or not.
> +                */
> +               if (host->platform->hardware_flush_all_cmds ||
> +                   cmd->data->flags & MMC_DATA_WRITE)
> +                       /* hardware flush: */
> +                       ictl |= MESON_SDHC_ICTL_DMA_DONE;
> +               else
> +                       /* software flush: */
> +                       ictl |= MESON_SDHC_ICTL_DATA_XFER_OK;
> +       } else {
> +               pack_len = 0;
> +
> +               ictl |= MESON_SDHC_ICTL_RESP_OK;
> +       }
> +
> +       if (cmd->opcode == MMC_STOP_TRANSMISSION)
> +               send |= MESON_SDHC_SEND_DATA_STOP;
> +
> +       if (cmd->flags & MMC_RSP_PRESENT)
> +               send |= MESON_SDHC_SEND_CMD_HAS_RESP;
> +
> +       if (cmd->flags & MMC_RSP_136) {
> +               send |= MESON_SDHC_SEND_RESP_LEN;
> +               send |= MESON_SDHC_SEND_RESP_NO_CRC;
> +       }
> +
> +       if (!(cmd->flags & MMC_RSP_CRC))
> +               send |= MESON_SDHC_SEND_RESP_NO_CRC;
> +
> +       if (cmd->flags & MMC_RSP_BUSY)
> +               send |= MESON_SDHC_SEND_R1B;
> +
> +       /* enable the new IRQs and mask all pending ones */
> +       regmap_write(host->regmap, MESON_SDHC_ICTL, ictl);
> +       regmap_write(host->regmap, MESON_SDHC_ISTA, MESON_SDHC_ISTA_ALL_IRQS);
> +
> +       regmap_write(host->regmap, MESON_SDHC_ARGU, cmd->arg);
> +
> +       regmap_update_bits(host->regmap, MESON_SDHC_CTRL,
> +                          MESON_SDHC_CTRL_PACK_LEN,
> +                          FIELD_PREP(MESON_SDHC_CTRL_PACK_LEN, pack_len));
> +
> +       if (cmd->data)
> +               regmap_write(host->regmap, MESON_SDHC_ADDR,
> +                            sg_dma_address(cmd->data->sg));
> +
> +       meson_mx_sdhc_wait_cmd_ready(mmc);
> +
> +       if (cmd->data)
> +               host->platform->set_pdma(mmc);
> +
> +       if (host->platform->wait_before_send)
> +               host->platform->wait_before_send(mmc);
> +
> +       regmap_write(host->regmap, MESON_SDHC_SEND, send);

Isn't there a configurable timeout to set for the command?

I mean the driver sets mmc->max_busy_timeout to 30s in ->probe(), but
can the timeout be configured to a lower value?

[...]

> +static void meson_mx_sdhc_wait_before_send_meson8(struct mmc_host *mmc)
> +{
> +       struct meson_mx_sdhc_host *host = mmc_priv(mmc);
> +       u32 val;
> +       int ret;
> +
> +       ret = regmap_read_poll_timeout(host->regmap, MESON_SDHC_ESTA, val,
> +                                      val == 0, 1, 200);

Please use defines for timeout values.

> +       if (ret)
> +               dev_warn(mmc_dev(mmc),
> +                        "Failed to wait for ESTA to clear: 0x%08x\n", val);
> +
> +       if (host->cmd->data && host->cmd->data->flags & MMC_DATA_WRITE) {
> +               ret = regmap_read_poll_timeout(host->regmap, MESON_SDHC_STAT,
> +                                              val,
> +                                              val & MESON_SDHC_STAT_TXFIFO_CNT,
> +                                              1, 200);

Ditto.

> +               if (ret)
> +                       dev_warn(mmc_dev(mmc),
> +                                "Failed to wait for TX FIFO to fill\n");
> +       }
> +}
> +

[...]

> +
> +static int meson_mx_sdhc_probe(struct platform_device *pdev)
> +{
> +       struct device *dev = &pdev->dev;
> +       struct meson_mx_sdhc_host *host;
> +       struct mmc_host *mmc;
> +       void __iomem *base;
> +       int ret, irq;
> +
> +       mmc = mmc_alloc_host(sizeof(*host), dev);
> +       if (!mmc)
> +               return -ENOMEM;
> +
> +       ret = devm_add_action_or_reset(dev, (void(*)(void *))mmc_free_host,
> +                                      mmc);
> +       if (ret) {
> +               dev_err(dev, "Failed to register mmc_free_host action\n");
> +               return ret;
> +       }
> +
> +       host = mmc_priv(mmc);
> +       host->mmc = mmc;
> +
> +       platform_set_drvdata(pdev, host);
> +
> +       host->platform = device_get_match_data(dev);
> +       if (!host->platform)
> +               return -EINVAL;
> +
> +       base = devm_platform_ioremap_resource(pdev, 0);
> +       if (IS_ERR(base))
> +               return PTR_ERR(base);
> +
> +       host->regmap = devm_regmap_init_mmio(dev, base,
> +                                            &meson_mx_sdhc_regmap_config);
> +       if (IS_ERR(host->regmap))
> +               return PTR_ERR(host->regmap);
> +
> +       host->pclk = devm_clk_get(dev, "pclk");
> +       if (IS_ERR(host->pclk))
> +               return PTR_ERR(host->pclk);
> +
> +       /* accessing any register requires the module clock to be enabled: */
> +       ret = clk_prepare_enable(host->pclk);
> +       if (ret) {
> +               dev_err(dev, "Failed to enable 'pclk' clock\n");
> +               return ret;
> +       }
> +
> +       meson_mx_sdhc_init_hw(mmc);
> +
> +       ret = meson_mx_sdhc_register_clkc(host);
> +       if (ret)
> +               goto err_disable_pclk;
> +
> +       host->bulk_clks[0].id = "mod_clk";
> +       host->bulk_clks[1].id = "sd_clk";
> +       host->bulk_clks[2].id = "tx_clk";
> +       host->bulk_clks[3].id = "rx_clk";
> +       ret = devm_clk_bulk_get(dev, MESON_SDHC_NUM_BULK_CLKS,
> +                               host->bulk_clks);
> +       if (ret)
> +               goto err_clkc_pdev_unregister;
> +
> +       host->sd_clk = host->bulk_clks[1].clk;
> +
> +       /* Get regulators and the supported OCR mask */
> +       ret = mmc_regulator_get_supply(mmc);
> +       if (ret)
> +               goto err_clkc_pdev_unregister;
> +
> +       mmc->max_req_size = SZ_128K;
> +       mmc->max_seg_size = mmc->max_req_size;
> +       mmc->max_blk_count = FIELD_GET(MESON_SDHC_SEND_TOTAL_PACK, ~0);
> +       mmc->max_blk_size = MESON_SDHC_MAX_BLK_SIZE;
> +       mmc->max_busy_timeout = 30 * MSEC_PER_SEC;
> +       mmc->f_min = clk_round_rate(host->sd_clk, 1);
> +       mmc->f_max = clk_round_rate(host->sd_clk, ULONG_MAX);
> +       mmc->max_current_180 = 300;
> +       mmc->max_current_330 = 300;
> +       mmc->caps |= MMC_CAP_ERASE | MMC_CAP_HW_RESET;

Should you also set MMC_CAP_WAIT_WHILE_BUSY? It sounded like the
driver supported this.

> +       mmc->ops = &meson_mx_sdhc_ops;
> +
> +       ret = mmc_of_parse(mmc);
> +       if (ret)
> +               goto err_clkc_pdev_unregister;
> +
> +       irq = platform_get_irq(pdev, 0);
> +       ret = devm_request_threaded_irq(dev, irq, meson_mx_sdhc_irq,
> +                                       meson_mx_sdhc_irq_thread, IRQF_ONESHOT,
> +                                       NULL, host);
> +       if (ret)
> +               goto err_clkc_pdev_unregister;
> +
> +       ret = mmc_add_host(mmc);
> +       if (ret)
> +               goto err_clkc_pdev_unregister;
> +
> +       return 0;
> +
> +err_clkc_pdev_unregister:
> +       platform_device_unregister(host->clkc_pdev);
> +err_disable_pclk:
> +       clk_disable_unprepare(host->pclk);
> +       return ret;
> +}
> +

[...]

FYI: I left out all comments related to the clock provider
initialization. I think it makes better sense to review that code,
after you have converted to use the devm_clk_hw_register() and avoid
registering a separate driver for it.

Other than the minor comments, this looks good to me.

Kind regards
Uffe