Re: [PATCH v9 2/2] spi: loongson: add bus driver for the loongson spi controller

[zhuyinbo]

在 2023/5/10 下午3:03, Andy Shevchenko 写道:
> _original_
>
>         const char rdiv[12] = {0, 1, 4, 2, 3, 5, 6, 7, 8, 9, 10, 11};
>
>                 div = DIV_ROUND_UP_ULL(loongson_spi->clk_rate, hz);
>                 if (div < 2)
>                         div = 2;
>                 if (div > 4096)
>                         div = 4096;
>
>                 bit = fls(div) - 1;
>                 if ((1<<bit) == div)
>                         bit--;
>                 div_tmp = rdiv[bit];
>
>                 loongson_spi->spcr = div_tmp & 3;
>                 loongson_spi->sper = (div_tmp >> 2) & 3;
>
> _proposed_
>
>         const char rdiv[12] = {0, 1, 4, 2, 3, 5, 6, 7, 8, 9, 10, 11};
>
> // as far as I understood the above table
> 00 00  [0]  <= 2
> 00 01  [1]  <= 4
> 01 00  [2]  <= 8
> 00 10  [3]  <= 16
> 00 11  [4]  <= 32
> 01 01  [5]  <= 64
> 01 10  [6]  <= 128
> 01 11  [7]  <= 256
> 10 00  [8]  <= 512
> 10 01  [9]  <= 1024
> 10 10  [10]  <= 2048
> 10 11  [11]  <= 4096
>
>                 div =
> clamp_val(DIV_ROUND_UP_ULL(loongson_spi->clk_rate, hz), 2, 4096);
>                 div_tmp = rdiv[fls(div - 1)];
>
>                 loongson_spi->spcr = (div_tmp & GENMASK(1, 0)) >> 0;
>                 loongson_spi->sper = (div_tmp & GENMASK(3, 2)) >> 2;
>
> But TBH I would expect the author to think about it more.


In previous code, if mode need to be updated but clk not to be updated
then the clk settting code will be also run.  so, I think it is better
to confiure clk and mode separately.

if (hz && loongson_spi->hz != hz) {
       div = clamp_val(DIV_ROUND_UP_ULL(loongson_spi->clk_rate, hz), 2,
                      4096);
       div_tmp = rdiv[fls(div - 1)];
       loongson_spi->spcr = (div_tmp & GENMASK(1, 0)) >> 0;
       loongson_spi->sper = (div_tmp & GENMASK(3, 2)) >> 2;
       val = loongson_spi_read_reg(loongson_spi, LOONGSON_SPI_SPCR_REG);
       loongson_spi_write_reg(loongson_spi, LOONGSON_SPI_SPCR_REG,
			(val & ~3) | loongson_spi->spcr);
       val = loongson_spi_read_reg(loongson_spi, LOONGSON_SPI_SPER_REG);
       loongson_spi_write_reg(loongson_spi, LOONGSON_SPI_SPER_REG,
                       (val & ~3) |loongson_spi->sper);
        loongson_spi->hz = hz;
}

if ((spi->mode ^ loongson_spi->mode) & SPI_MODE_X_MASK) {
       val = loongson_spi_read_reg(loongson_spi, LOONGSON_SPI_SPCR_REG);
       val &= ~(LOONGSON_SPI_SPCR_CPOL | LOONGSON_SPI_SPCR_CPHA);
       if (spi->mode & SPI_CPOL)
		val |= LOONGSON_SPI_SPCR_CPOL;
       if (spi->mode & SPI_CPHA)
		val |= LOONGSON_SPI_SPCR_CPHA;
       loongson_spi_write_reg(loongson_spi, LOONGSON_SPI_SPCR_REG, val);
       loongson_spi->mode |= spi->mode;
}

> Also the check can be modified to have less indented code:
>
>         if (hz && loongson_spi->hz == hz)
>            return 0;
>
>         if (!((spi->mode ^ loongson_spi->mode) & SPI_MODE_X_MASK))
>            return 0;

The setting register about clk and mode was the same SPCR register, so
only the clk and mode don't need to be updated in the same, then return 
0, so the code seems to be as follows. But setting clk and mode 
separately doesn't require follows judgement.

        if ((hz && loongson_spi->hz == hz) &&
            !((spi->mode ^ loongson_spi->mode) & SPI_MODE_X_MASK))
                return 0;

Thanks.

>