Re: [PATCH 1/1] DaVinci: MMC: V4: MMC/SD controller driver for DaVinci family.

[Pierre Ossman <drzeus-mmc@drzeus.cx>]

On Thu,  2 Apr 2009 15:26:36 +0530
Purushotam Kumar <purushotam@ti.com> wrote:

>  arch/arm/mach-davinci/include/mach/mmc.h |   25 +
>  drivers/mmc/host/Kconfig                 |    8 +
>  drivers/mmc/host/Makefile                |    1 +
>  drivers/mmc/host/davinci_mmc.c           | 1264 ++++++++++++++++++++++++++++++
>  4 files changed, 1298 insertions(+), 0 deletions(-)
>  create mode 100644 arch/arm/mach-davinci/include/mach/mmc.h
>  create mode 100644 drivers/mmc/host/davinci_mmc.c

I want a MAINTAINERS entry for this driver as well.

> +	/*
> +	 * Before non-DMA WRITE commands the controller needs priming:
> +	 * FIFO should be populated with 32 bytes i.e. whatever is the FIFO size
> +	 */
> +	if (!host->do_dma && (host->data_dir == DAVINCI_MMC_DATADIR_WRITE) &&
> +		((cmd->opcode == MMC_WRITE_BLOCK) ||
> +		(cmd->opcode == MMC_WRITE_MULTIPLE_BLOCK)))
> +		davinci_fifo_data_trans(host, DAVINCI_MMC_FIFO_SIZE_BYTE);

I'm still not following the requirements here. Why would the hardware
only need to have the FIFO primed for those two commands and not every
kind of write?

> +
> +	/* We know sg_len and ccnt will never be out of range because
> +	 * we told the mmc layer which in turn tells the block layer
> +	 * to ensure that it only hands us one scatterlist segment
> +	 * per EDMA PARAM entry.  Update the PARAM
> +	 * entries needed for each segment of this scatterlist.
> +	 */
> +	for (slot = channel, link = 0, sg = data->sg, sg_len = host->sg_len;
> +			sg_len-- != 0 && bytes_left;
> +			sg = sg_next(sg), slot = host->links[link++]) {

You might be able to make this cleaner using the for_each_sg helper.

> +static unsigned int calculate_freq_for_card(struct mmc_davinci_host *host,
> +	unsigned int mmc_req_freq)
> +{
> +	unsigned int mmc_freq = 0, cpu_arm_clk = 0, mmc_push_pull = 0;
> +
> +	cpu_arm_clk = host->mmc_input_clk;
> +	if (mmc_req_freq && cpu_arm_clk > (2 * mmc_req_freq))
> +		mmc_push_pull = ((unsigned int)cpu_arm_clk
> +				/ (2 * mmc_req_freq)) - 1;
> +	else
> +		mmc_push_pull = 0;
> +
> +	mmc_freq = (unsigned int)cpu_arm_clk / (2 * (mmc_push_pull + 1));
> +
> +	if (mmc_freq > mmc_req_freq)
> +		mmc_push_pull = mmc_push_pull + 1;

The naming of the divider is a bit confusing here. :)

> +
> +	/* Convert ns to clock cycles */
> +	if (mmc_req_freq < 400000)
> +		ns_in_one_cycle = 1000000 / ((cpu_arm_clk
> +					/ (2 * (mmc_push_pull + 1)))/1000);
> +	else
> +		ns_in_one_cycle = 1000000 / ((cpu_arm_clk
> +					/ (2 * (mmc_push_pull + 1)))/1000000);
> +

The clock frequency is per host, so store this in the mmc_davinci_host
structure.

> +	if (ios->bus_mode == MMC_BUSMODE_OPENDRAIN) {
> +		u32 temp;
> +
> +		/* Ignoring the init clock value passed for fixing the inter
> +		 * operability with different cards.
> +		 */
> +		open_drain_freq = ((unsigned int)cpu_arm_clk
> +				/ (2 * MMCSD_INIT_CLOCK)) - 1;
> +
> +		if (open_drain_freq > 0xFF)
> +			open_drain_freq = 0xFF;
> +
> +		temp = readl(host->base + DAVINCI_MMCCLK) & ~MMCCLK_CLKRT_MASK;
> +		temp |= open_drain_freq;
> +		writel(temp, host->base + DAVINCI_MMCCLK);
> +
> +		/* Convert ns to clock cycles */
> +		ns_in_one_cycle = (1000000) / (MMCSD_INIT_CLOCK/1000);

I'm still not sure why you need this. Open drain mode is always used at
a low frequency anyway.

Rgds
-- 
     -- Pierre Ossman

  Linux kernel, MMC maintainer        http://www.kernel.org
  rdesktop, core developer          http://www.rdesktop.org
  TigerVNC, core developer          http://www.tigervnc.org

  WARNING: This correspondence is being monitored by the
  Swedish government. Make sure your server uses encryption
  for SMTP traffic and consider using PGP for end-to-end
  encryption.