A hardcoded reference clock of 48 MHz is used to calculate the
clock divisor values, but the reference clock frequency can be
different across devices and can be configured which can cause
a mismatch between the reported frequency and actual SPI clock
frequency observed. Fix this by fetching the clock rate from
the clock provider and falling back to hardcoded reference only
if the clock is not supplied.
Fixes: 2cd7d393f461 ("arm64: dts: ti: k3-am654: Add McSPI DT nodes")
Signed-off-by: Vaishnav Achath <[email protected]>
---
V1->V2:
* Use u32 instead of unsigned int
* rename clk to ref_clk and rebase for latest struct ctlr rename changes.
Blamed commit is the first device where the default reference clock was
different from the hardcoded value.
Tested on TDA4VM SK (6.6.0-rc3-next-20230926)
https://gist.github.com/vaishnavachath/7bbc6dd6e02294aed3e8c547cfa039c2
V1 - https://lore.kernel.org/all/[email protected]/
drivers/spi/spi-omap2-mcspi.c | 31 ++++++++++++++++++++-----------
1 file changed, 20 insertions(+), 11 deletions(-)
diff --git a/drivers/spi/spi-omap2-mcspi.c b/drivers/spi/spi-omap2-mcspi.c
index 79888e6c54c2..ddf1c684bcc7 100644
--- a/drivers/spi/spi-omap2-mcspi.c
+++ b/drivers/spi/spi-omap2-mcspi.c
@@ -125,10 +125,12 @@ struct omap2_mcspi {
struct omap2_mcspi_dma *dma_channels;
struct device *dev;
struct omap2_mcspi_regs ctx;
+ struct clk *ref_clk;
int fifo_depth;
bool target_aborted;
unsigned int pin_dir:1;
size_t max_xfer_len;
+ u32 ref_clk_hz;
};
struct omap2_mcspi_cs {
@@ -880,12 +882,12 @@ omap2_mcspi_txrx_pio(struct spi_device *spi, struct spi_transfer *xfer)
return count - c;
}
-static u32 omap2_mcspi_calc_divisor(u32 speed_hz)
+static u32 omap2_mcspi_calc_divisor(u32 speed_hz, u32 ref_clk_hz)
{
u32 div;
for (div = 0; div < 15; div++)
- if (speed_hz >= (OMAP2_MCSPI_MAX_FREQ >> div))
+ if (speed_hz >= (ref_clk_hz >> div))
return div;
return 15;
@@ -897,7 +899,7 @@ static int omap2_mcspi_setup_transfer(struct spi_device *spi,
{
struct omap2_mcspi_cs *cs = spi->controller_state;
struct omap2_mcspi *mcspi;
- u32 l = 0, clkd = 0, div, extclk = 0, clkg = 0;
+ u32 ref_clk_hz, l = 0, clkd = 0, div, extclk = 0, clkg = 0;
u8 word_len = spi->bits_per_word;
u32 speed_hz = spi->max_speed_hz;
@@ -911,14 +913,15 @@ static int omap2_mcspi_setup_transfer(struct spi_device *spi,
if (t && t->speed_hz)
speed_hz = t->speed_hz;
- speed_hz = min_t(u32, speed_hz, OMAP2_MCSPI_MAX_FREQ);
- if (speed_hz < (OMAP2_MCSPI_MAX_FREQ / OMAP2_MCSPI_MAX_DIVIDER)) {
- clkd = omap2_mcspi_calc_divisor(speed_hz);
- speed_hz = OMAP2_MCSPI_MAX_FREQ >> clkd;
+ ref_clk_hz = mcspi->ref_clk_hz;
+ speed_hz = min_t(u32, speed_hz, ref_clk_hz);
+ if (speed_hz < (ref_clk_hz / OMAP2_MCSPI_MAX_DIVIDER)) {
+ clkd = omap2_mcspi_calc_divisor(speed_hz, ref_clk_hz);
+ speed_hz = ref_clk_hz >> clkd;
clkg = 0;
} else {
- div = (OMAP2_MCSPI_MAX_FREQ + speed_hz - 1) / speed_hz;
- speed_hz = OMAP2_MCSPI_MAX_FREQ / div;
+ div = (ref_clk_hz + speed_hz - 1) / speed_hz;
+ speed_hz = ref_clk_hz / div;
clkd = (div - 1) & 0xf;
extclk = (div - 1) >> 4;
clkg = OMAP2_MCSPI_CHCONF_CLKG;
@@ -1448,8 +1451,6 @@ static int omap2_mcspi_probe(struct platform_device *pdev)
ctlr->cleanup = omap2_mcspi_cleanup;
ctlr->target_abort = omap2_mcspi_target_abort;
ctlr->dev.of_node = node;
- ctlr->max_speed_hz = OMAP2_MCSPI_MAX_FREQ;
- ctlr->min_speed_hz = OMAP2_MCSPI_MAX_FREQ >> 15;
ctlr->use_gpio_descriptors = true;
platform_set_drvdata(pdev, ctlr);
@@ -1519,6 +1520,14 @@ static int omap2_mcspi_probe(struct platform_device *pdev)
goto free_ctlr;
}
+ mcspi->ref_clk = devm_clk_get_optional_enabled(&pdev->dev, NULL);
+ if (mcspi->ref_clk)
+ mcspi->ref_clk_hz = clk_get_rate(mcspi->ref_clk);
+ else
+ mcspi->ref_clk_hz = OMAP2_MCSPI_MAX_FREQ;
+ ctlr->max_speed_hz = mcspi->ref_clk_hz;
+ ctlr->min_speed_hz = mcspi->ref_clk_hz >> 15;
+
pm_runtime_use_autosuspend(&pdev->dev);
pm_runtime_set_autosuspend_delay(&pdev->dev, SPI_AUTOSUSPEND_TIMEOUT);
pm_runtime_enable(&pdev->dev);
--
2.17.1
On Sep 26, 2023 at 17:08:12 +0530, Vaishnav Achath wrote:
> A hardcoded reference clock of 48 MHz is used to calculate the
> clock divisor values, but the reference clock frequency can be
> different across devices and can be configured which can cause
> a mismatch between the reported frequency and actual SPI clock
> frequency observed. Fix this by fetching the clock rate from
> the clock provider and falling back to hardcoded reference only
> if the clock is not supplied.
>
> Fixes: 2cd7d393f461 ("arm64: dts: ti: k3-am654: Add McSPI DT nodes")
>
> Signed-off-by: Vaishnav Achath <[email protected]>
> ---
>
> V1->V2:
> * Use u32 instead of unsigned int
> * rename clk to ref_clk and rebase for latest struct ctlr rename changes.
>
> Blamed commit is the first device where the default reference clock was
> different from the hardcoded value.
> Tested on TDA4VM SK (6.6.0-rc3-next-20230926)
> https://gist.github.com/vaishnavachath/7bbc6dd6e02294aed3e8c547cfa039c2
Superb, the logic analyzer snapshot in the end makes it so much cooler
hehe :)
>
> V1 - https://lore.kernel.org/all/[email protected]/
>
> drivers/spi/spi-omap2-mcspi.c | 31 ++++++++++++++++++++-----------
> 1 file changed, 20 insertions(+), 11 deletions(-)
>
> diff --git a/drivers/spi/spi-omap2-mcspi.c b/drivers/spi/spi-omap2-mcspi.c
> index 79888e6c54c2..ddf1c684bcc7 100644
> --- a/drivers/spi/spi-omap2-mcspi.c
> +++ b/drivers/spi/spi-omap2-mcspi.c
> @@ -125,10 +125,12 @@ struct omap2_mcspi {
> struct omap2_mcspi_dma *dma_channels;
> struct device *dev;
> struct omap2_mcspi_regs ctx;
> + struct clk *ref_clk;
> int fifo_depth;
> bool target_aborted;
> unsigned int pin_dir:1;
> size_t max_xfer_len;
> + u32 ref_clk_hz;
Looks good now.
> };
>
> struct omap2_mcspi_cs {
> @@ -880,12 +882,12 @@ omap2_mcspi_txrx_pio(struct spi_device *spi, struct spi_transfer *xfer)
> return count - c;
> }
>
> -static u32 omap2_mcspi_calc_divisor(u32 speed_hz)
> +static u32 omap2_mcspi_calc_divisor(u32 speed_hz, u32 ref_clk_hz)
> {
> u32 div;
>
> for (div = 0; div < 15; div++)
> - if (speed_hz >= (OMAP2_MCSPI_MAX_FREQ >> div))
> + if (speed_hz >= (ref_clk_hz >> div))
> return div;
>
> return 15;
> @@ -897,7 +899,7 @@ static int omap2_mcspi_setup_transfer(struct spi_device *spi,
> {
> struct omap2_mcspi_cs *cs = spi->controller_state;
> struct omap2_mcspi *mcspi;
> - u32 l = 0, clkd = 0, div, extclk = 0, clkg = 0;
> + u32 ref_clk_hz, l = 0, clkd = 0, div, extclk = 0, clkg = 0;
> u8 word_len = spi->bits_per_word;
> u32 speed_hz = spi->max_speed_hz;
>
> @@ -911,14 +913,15 @@ static int omap2_mcspi_setup_transfer(struct spi_device *spi,
> if (t && t->speed_hz)
> speed_hz = t->speed_hz;
>
> - speed_hz = min_t(u32, speed_hz, OMAP2_MCSPI_MAX_FREQ);
> - if (speed_hz < (OMAP2_MCSPI_MAX_FREQ / OMAP2_MCSPI_MAX_DIVIDER)) {
> - clkd = omap2_mcspi_calc_divisor(speed_hz);
> - speed_hz = OMAP2_MCSPI_MAX_FREQ >> clkd;
> + ref_clk_hz = mcspi->ref_clk_hz;
> + speed_hz = min_t(u32, speed_hz, ref_clk_hz);
> + if (speed_hz < (ref_clk_hz / OMAP2_MCSPI_MAX_DIVIDER)) {
> + clkd = omap2_mcspi_calc_divisor(speed_hz, ref_clk_hz);
> + speed_hz = ref_clk_hz >> clkd;
> clkg = 0;
> } else {
> - div = (OMAP2_MCSPI_MAX_FREQ + speed_hz - 1) / speed_hz;
> - speed_hz = OMAP2_MCSPI_MAX_FREQ / div;
> + div = (ref_clk_hz + speed_hz - 1) / speed_hz;
> + speed_hz = ref_clk_hz / div;
Looks good.
> clkd = (div - 1) & 0xf;
> extclk = (div - 1) >> 4;
> clkg = OMAP2_MCSPI_CHCONF_CLKG;
> @@ -1448,8 +1451,6 @@ static int omap2_mcspi_probe(struct platform_device *pdev)
> ctlr->cleanup = omap2_mcspi_cleanup;
> ctlr->target_abort = omap2_mcspi_target_abort;
> ctlr->dev.of_node = node;
> - ctlr->max_speed_hz = OMAP2_MCSPI_MAX_FREQ;
> - ctlr->min_speed_hz = OMAP2_MCSPI_MAX_FREQ >> 15;
> ctlr->use_gpio_descriptors = true;
>
> platform_set_drvdata(pdev, ctlr);
> @@ -1519,6 +1520,14 @@ static int omap2_mcspi_probe(struct platform_device *pdev)
> goto free_ctlr;
> }
>
> + mcspi->ref_clk = devm_clk_get_optional_enabled(&pdev->dev, NULL);
> + if (mcspi->ref_clk)
> + mcspi->ref_clk_hz = clk_get_rate(mcspi->ref_clk);
> + else
> + mcspi->ref_clk_hz = OMAP2_MCSPI_MAX_FREQ;
> + ctlr->max_speed_hz = mcspi->ref_clk_hz;
> + ctlr->min_speed_hz = mcspi->ref_clk_hz >> 15;
nit: Can we improve on previous code by making 15 into a macro?
> +
> pm_runtime_use_autosuspend(&pdev->dev);
> pm_runtime_set_autosuspend_delay(&pdev->dev, SPI_AUTOSUSPEND_TIMEOUT);
> pm_runtime_enable(&pdev->dev);
> --
> 2.17.1
>
Thanks for addressing comments,
Reviewed-by: Dhruva Gole <[email protected]>
--
Best regards,
Dhruva Gole <[email protected]>