[PATCH 4/8] sdhci: sdhci-esdhci-imx: add sd3.0 clock tuning support
Ulf Hansson
ulf.hansson at linaro.org
Mon Sep 16 03:48:11 EDT 2013
On 5 September 2013 19:52, Dong Aisheng <dongas86 at gmail.com> wrote:
> On Thu, Sep 5, 2013 at 3:33 PM, Ulf Hansson <ulf.hansson at linaro.org> wrote:
>> On 4 September 2013 14:54, Dong Aisheng <b29396 at freescale.com> wrote:
>>> Freescale i.MX6Q/DL uSDHC clock tuning progress is a little different from
>>> the standard tuning process defined in host controller spec v3.0.
>>> Thus we use platform_execute_tuning instead of standard sdhci tuning.
>>>
>>> The main difference are:
>>> 1) not only generate Buffer Read Ready interrupt when tuning is performing.
>>> It generates all other DATA interrupts like the normal data command.
>>> 2) SDHCI_CTRL_EXEC_TUNING is not automatically cleared by HW,
>>> instead it's controlled by SW.
>>> 3) SDHCI_CTRL_TUNED_CLK is not automatically set by HW,
>>> it's controlled by SW.
>>> 4) the clock delay for every tuning is set by SW.
>>>
>>> Signed-off-by: Dong Aisheng <b29396 at freescale.com>
>>> ---
>>> drivers/mmc/host/sdhci-esdhc-imx.c | 194 +++++++++++++++++++++++++++++++++++-
>>> 1 files changed, 193 insertions(+), 1 deletions(-)
>>>
>>> diff --git a/drivers/mmc/host/sdhci-esdhc-imx.c b/drivers/mmc/host/sdhci-esdhc-imx.c
>>> index 3118a82..36b9f63 100644
>>> --- a/drivers/mmc/host/sdhci-esdhc-imx.c
>>> +++ b/drivers/mmc/host/sdhci-esdhc-imx.c
>>> @@ -34,9 +34,21 @@
>>> #define ESDHC_WTMK_LVL 0x44
>>> #define ESDHC_MIX_CTRL 0x48
>>> #define ESDHC_MIX_CTRL_AC23EN (1 << 7)
>>> +#define ESDHC_MIX_CTRL_EXE_TUNE (1 << 22)
>>> +#define ESDHC_MIX_CTRL_SMPCLK_SEL (1 << 23)
>>> +#define ESDHC_MIX_CTRL_AUTO_TUNE (1 << 24)
>>> +#define ESDHC_MIX_CTRL_FBCLK_SEL (1 << 25)
>>> /* Bits 3 and 6 are not SDHCI standard definitions */
>>> #define ESDHC_MIX_CTRL_SDHCI_MASK 0xb7
>>>
>>> +/* tune control register */
>>> +#define ESDHC_TUNE_CTRL_STATUS 0x68
>>> +#define ESDHC_TUNE_CTRL_STEP 1
>>> +#define ESDHC_TUNE_CTRL_MIN 0
>>> +#define ESDHC_TUNE_CTRL_MAX ((1 << 7) - 1)
>>> +
>>> +#define ESDHC_TUNING_BLOCK_PATTERN_LEN 64
>>> +
>>> /*
>>> * Our interpretation of the SDHCI_HOST_CONTROL register
>>> */
>>> @@ -87,7 +99,7 @@ struct pltfm_imx_data {
>>> MULTIBLK_IN_PROCESS, /* exact multiblock cmd in process */
>>> WAIT_FOR_INT, /* sent CMD12, waiting for response INT */
>>> } multiblock_status;
>>> -
>>> + u32 uhs_mode;
>>> };
>>>
>>> static struct platform_device_id imx_esdhc_devtype[] = {
>>> @@ -161,6 +173,17 @@ static u32 esdhc_readl_le(struct sdhci_host *host, int reg)
>>> struct pltfm_imx_data *imx_data = pltfm_host->priv;
>>> u32 val = readl(host->ioaddr + reg);
>>>
>>> + if (unlikely(reg == SDHCI_PRESENT_STATE)) {
>>> + u32 fsl_prss = val;
>>> + val = 0;
>>> + /* save the least 20 bits */
>>> + val |= fsl_prss & 0x000FFFFF;
>>> + /* move dat[0-3] bits */
>>> + val |= (fsl_prss & 0x0F000000) >> 4;
>>> + /* move cmd line bit */
>>> + val |= (fsl_prss & 0x00800000) << 1;
>>> + }
>>> +
>>> if (unlikely(reg == SDHCI_CAPABILITIES)) {
>>> /* In FSL esdhc IC module, only bit20 is used to indicate the
>>> * ADMA2 capability of esdhc, but this bit is messed up on
>>> @@ -175,6 +198,17 @@ static u32 esdhc_readl_le(struct sdhci_host *host, int reg)
>>> }
>>> }
>>>
>>> + if (unlikely(reg == SDHCI_CAPABILITIES_1) && is_imx6q_usdhc(imx_data))
>>> + val = SDHCI_SUPPORT_DDR50 | SDHCI_SUPPORT_SDR104
>>> + | SDHCI_SUPPORT_SDR50;
>>> +
>>> + if (unlikely(reg == SDHCI_MAX_CURRENT) && is_imx6q_usdhc(imx_data)) {
>>> + val = 0;
>>> + val |= 0xFF << SDHCI_MAX_CURRENT_330_SHIFT;
>>> + val |= 0xFF << SDHCI_MAX_CURRENT_300_SHIFT;
>>> + val |= 0xFF << SDHCI_MAX_CURRENT_180_SHIFT;
>>> + }
>>> +
>>> if (unlikely(reg == SDHCI_INT_STATUS)) {
>>> if (val & ESDHC_INT_VENDOR_SPEC_DMA_ERR) {
>>> val &= ~ESDHC_INT_VENDOR_SPEC_DMA_ERR;
>>> @@ -253,6 +287,8 @@ static u16 esdhc_readw_le(struct sdhci_host *host, int reg)
>>> {
>>> struct sdhci_pltfm_host *pltfm_host = sdhci_priv(host);
>>> struct pltfm_imx_data *imx_data = pltfm_host->priv;
>>> + u16 ret = 0;
>>> + u32 val;
>>>
>>> if (unlikely(reg == SDHCI_HOST_VERSION)) {
>>> reg ^= 2;
>>> @@ -265,6 +301,25 @@ static u16 esdhc_readw_le(struct sdhci_host *host, int reg)
>>> }
>>> }
>>>
>>> + if (unlikely(reg == SDHCI_HOST_CONTROL2)) {
>>> + val = readl(host->ioaddr + ESDHC_VENDOR_SPEC);
>>> + if (val & ESDHC_VENDOR_SPEC_VSELECT)
>>> + ret |= SDHCI_CTRL_VDD_180;
>>> +
>>> + if (is_imx6q_usdhc(imx_data)) {
>>> + val = readl(host->ioaddr + ESDHC_MIX_CTRL);
>>> + if (val & ESDHC_MIX_CTRL_EXE_TUNE)
>>> + ret |= SDHCI_CTRL_EXEC_TUNING;
>>> + if (val & ESDHC_MIX_CTRL_SMPCLK_SEL)
>>> + ret |= SDHCI_CTRL_TUNED_CLK;
>>> + }
>>> +
>>> + ret |= (imx_data->uhs_mode & SDHCI_CTRL_UHS_MASK);
>>> + ret &= ~SDHCI_CTRL_PRESET_VAL_ENABLE;
>>> +
>>> + return ret;
>>> + }
>>> +
>>> return readw(host->ioaddr + reg);
>>> }
>>>
>>> @@ -272,8 +327,32 @@ static void esdhc_writew_le(struct sdhci_host *host, u16 val, int reg)
>>> {
>>> struct sdhci_pltfm_host *pltfm_host = sdhci_priv(host);
>>> struct pltfm_imx_data *imx_data = pltfm_host->priv;
>>> + u32 new_val = 0;
>>>
>>> switch (reg) {
>>> + case SDHCI_CLOCK_CONTROL:
>>> + new_val = readl(host->ioaddr + ESDHC_VENDOR_SPEC);
>>> + if (val & SDHCI_CLOCK_CARD_EN)
>>> + new_val |= ESDHC_VENDOR_SPEC_FRC_SDCLK_ON;
>>> + else
>>> + new_val &= ~ESDHC_VENDOR_SPEC_FRC_SDCLK_ON;
>>> + writel(new_val, host->ioaddr + ESDHC_VENDOR_SPEC);
>>> + return;
>>> + case SDHCI_HOST_CONTROL2:
>>> + new_val = readl(host->ioaddr + ESDHC_VENDOR_SPEC);
>>> + if (val & SDHCI_CTRL_VDD_180)
>>> + new_val |= ESDHC_VENDOR_SPEC_VSELECT;
>>> + else
>>> + new_val &= ~ESDHC_VENDOR_SPEC_VSELECT;
>>> + writel(new_val, host->ioaddr + ESDHC_VENDOR_SPEC);
>>> + imx_data->uhs_mode = val & SDHCI_CTRL_UHS_MASK;
>>> + new_val = readl(host->ioaddr + ESDHC_MIX_CTRL);
>>> + if (val & SDHCI_CTRL_TUNED_CLK)
>>> + new_val |= ESDHC_MIX_CTRL_SMPCLK_SEL;
>>> + else
>>> + new_val &= ~ESDHC_MIX_CTRL_SMPCLK_SEL;
>>> + writel(new_val , host->ioaddr + ESDHC_MIX_CTRL);
>>> + return;
>>> case SDHCI_TRANSFER_MODE:
>>> if ((imx_data->flags & ESDHC_FLAG_MULTIBLK_NO_INT)
>>> && (host->cmd->opcode == SD_IO_RW_EXTENDED)
>>> @@ -451,6 +530,118 @@ static int esdhc_pltfm_bus_width(struct sdhci_host *host, int width)
>>> return 0;
>>> }
>>>
>>> +static void esdhc_prepare_tuning(struct sdhci_host *host, u32 val)
>>> +{
>>> + u32 reg;
>>> +
>>> + reg = readl(host->ioaddr + ESDHC_MIX_CTRL);
>>> + reg |= ESDHC_MIX_CTRL_EXE_TUNE | ESDHC_MIX_CTRL_SMPCLK_SEL |
>>> + ESDHC_MIX_CTRL_FBCLK_SEL;
>>> + writel(reg, host->ioaddr + ESDHC_MIX_CTRL);
>>> + writel((val << 8), host->ioaddr + ESDHC_TUNE_CTRL_STATUS);
>>> + dev_dbg(mmc_dev(host->mmc),
>>> + "tunning with delay 0x%x ESDHC_TUNE_CTRL_STATUS 0x%x\n",
>>> + val, readl(host->ioaddr + ESDHC_TUNE_CTRL_STATUS));
>>> +}
>>> +
>>> +static void request_done(struct mmc_request *mrq)
>>> +{
>>> + complete(&mrq->completion);
>>> +}
>>> +
>>> +static int esdhc_send_tuning_cmd(struct sdhci_host *host, u32 opcode)
>>
>> This function implements protocol related code. It should not reside
>> in a host driver but instead as an API in the mmc protocol layer which
>> host drivers shall call.
>>
>> I realize that there are already other host drivers implementing
>> similar protocol code as here. Those should move to use the new API
>> once it is available.
>
> That sounds like a good idea.
> The tuning comand sending process seems could be implemented in protocal layer
> since it's standard.
> One problem is that since the tuning command handling is tightly
> related to host controller.
> It may be a big rework on current code if implement such API and
> switch all the drivers to call it.
> And it may also increase the host driver complication since it needs
> to treat tuning
> command differently in driver and do specfic settings at specific
> places, e.g sdhci.
> Not sure it's good enough.
Hi Dong,
I am not proposing to move the entire tuning sequence to the protocol
layer, only the part that is related to send the actual tuning
command.
Typically, "esdhc_send_tuning_cmd" could be an API that host driver's
can call. I think this could prevent us from having a lot of
duplicated code in host drivers.
Not directly related to this patch; regarding periodic re-tuning; this
kind of feature should also be implemented in the protocol layer.
Otherwise each an every driver will re-implement similar
timer/work-queue and maybe even block statistics to trigger re-tuning.
Sdhci has already done this, which to me feels plain wrong for host
driver to do.
So before this goes out of control, we need to be aligned on what code
should reside the protocol layer, because that is in the end what this
discussion boils done to.
Kind regards
Ulf Hansson
>
> For imx, since it's based on sdhci, it may be less impact even once there's
> such an API for it to switch to.
> So i may would like to keep the using as sdhci currently if could,
> which also will block
> my following eMMC4.5 and SDIO3.0 work.
>
> Regards
> Dong Aisheng
>
>>
>> Kind regards
>> Ulf Hansson
>>
>>> +{
>>> + struct mmc_command cmd = {0};
>>> + struct mmc_request mrq = {0};
>>> + struct mmc_data data = {0};
>>> + struct scatterlist sg;
>>> + char tuning_pattern[ESDHC_TUNING_BLOCK_PATTERN_LEN];
>>> +
>>> + cmd.opcode = opcode;
>>> + cmd.arg = 0;
>>> + cmd.flags = MMC_RSP_R1 | MMC_CMD_ADTC;
>>> +
>>> + data.blksz = ESDHC_TUNING_BLOCK_PATTERN_LEN;
>>> + data.blocks = 1;
>>> + data.flags = MMC_DATA_READ;
>>> + data.sg = &sg;
>>> + data.sg_len = 1;
>>> +
>>> + sg_init_one(&sg, tuning_pattern, sizeof(tuning_pattern));
>>> +
>>> + mrq.cmd = &cmd;
>>> + mrq.cmd->mrq = &mrq;
>>> + mrq.data = &data;
>>> + mrq.data->mrq = &mrq;
>>> + mrq.cmd->data = mrq.data;
>>> +
>>> + mrq.done = request_done;
>>> + init_completion(&(mrq.completion));
>>> +
>>> + disable_irq(host->irq);
>>> + spin_lock(&host->lock);
>>> + host->mrq = &mrq;
>>> +
>>> + sdhci_send_command(host, mrq.cmd);
>>> +
>>> + spin_unlock(&host->lock);
>>> + enable_irq(host->irq);
>>> +
>>> + wait_for_completion(&mrq.completion);
>>> +
>>> + if (cmd.error)
>>> + return cmd.error;
>>> + if (data.error)
>>> + return data.error;
>>> +
>>> + return 0;
>>> +}
>>> +
>>> +static void esdhc_post_tuning(struct sdhci_host *host)
>>> +{
>>> + u32 reg;
>>> +
>>> + reg = readl(host->ioaddr + ESDHC_MIX_CTRL);
>>> + reg &= ~ESDHC_MIX_CTRL_EXE_TUNE;
>>> + writel(reg, host->ioaddr + ESDHC_MIX_CTRL);
>>> +}
>>> +
>>> +static int esdhc_executing_tuning(struct sdhci_host *host, u32 opcode)
>>> +{
>>> + int min, max, avg, ret;
>>> +
>>> + /* find the mininum delay first which can pass tuning*/
>>> + min = ESDHC_TUNE_CTRL_MIN;
>>> + while (min < ESDHC_TUNE_CTRL_MAX) {
>>> + esdhc_prepare_tuning(host, min);
>>> + if (!esdhc_send_tuning_cmd(host, opcode))
>>> + break;
>>> + min += ESDHC_TUNE_CTRL_STEP;
>>> + }
>>> +
>>> + /* find the maxinum delay which can not pass tuning*/
>>> + max = min + ESDHC_TUNE_CTRL_STEP;
>>> + while (max < ESDHC_TUNE_CTRL_MAX) {
>>> + esdhc_prepare_tuning(host, max);
>>> + if (esdhc_send_tuning_cmd(host, opcode)) {
>>> + max -= ESDHC_TUNE_CTRL_STEP;
>>> + break;
>>> + }
>>> + max += ESDHC_TUNE_CTRL_STEP;
>>> + }
>>> +
>>> + /* use average delay to get the best timing */
>>> + avg = (min + max) / 2;
>>> + esdhc_prepare_tuning(host, avg);
>>> + ret = esdhc_send_tuning_cmd(host, opcode);
>>> + esdhc_post_tuning(host);
>>> +
>>> + dev_dbg(mmc_dev(host->mmc), "tunning %s at 0x%x ret %d\n",
>>> + ret ? "failed" : "passed", avg, ret);
>>> +
>>> + return ret;
>>> +}
>>> +
>>> static const struct sdhci_ops sdhci_esdhc_ops = {
>>> .read_l = esdhc_readl_le,
>>> .read_w = esdhc_readw_le,
>>> @@ -462,6 +653,7 @@ static const struct sdhci_ops sdhci_esdhc_ops = {
>>> .get_min_clock = esdhc_pltfm_get_min_clock,
>>> .get_ro = esdhc_pltfm_get_ro,
>>> .platform_bus_width = esdhc_pltfm_bus_width,
>>> + .platform_execute_tuning = esdhc_executing_tuning,
>>> };
>>>
>>> static const struct sdhci_pltfm_data sdhci_esdhc_imx_pdata = {
>>> --
>>> 1.7.1
>>>
>>>
>>> --
>>> To unsubscribe from this list: send the line "unsubscribe linux-mmc" in
>>> the body of a message to majordomo at vger.kernel.org
>>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>>
>> _______________________________________________
>> linux-arm-kernel mailing list
>> linux-arm-kernel at lists.infradead.org
>> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
More information about the linux-arm-kernel
mailing list