[alsa-devel] [PATCH 4/4] pxa-ssp: switch from network mode to psp

[pHilipp Zabel]

On Wed, Mar 4, 2009 at 9:17 PM, Daniel Mack <daniel at caiaq.de> wrote:
> This patch uses the SSP's PSP functionality to provide I2S timings. The
> particular problem is that even though the datasheets state it should be
> possbible, there is no mode which uses the network feature with its
> associated time slots in a sane way to do what we need.
>
> Hence, in order to have full 64 bit I2S on the wire, we need to fiddle
> around with the SSP and the timing paramters a lot. There are some
> constants left in the code which can't be replaced by names because the
> true meaning of their registers remains nebulous.
>
> Signed-off-by: Daniel Mack <daniel at caiaq.de>
> Signed-off-by: Tim Ruetz <tim at caiaq.de>
> Cc: Mark Brown <broonie at opensource.wolfsonmicro.com>
> Cc: Liam Girdwood <lrg at kernel.org>
> Cc: Philipp Zabel <philipp.zabel at gmail.com>
> ---
>  sound/soc/pxa/pxa-ssp.c |   39 +++++++++++++++++++++------------------
>  1 files changed, 21 insertions(+), 18 deletions(-)
>
> diff --git a/sound/soc/pxa/pxa-ssp.c b/sound/soc/pxa/pxa-ssp.c
> index 45fb600..97f11d6 100644
> --- a/sound/soc/pxa/pxa-ssp.c
> +++ b/sound/soc/pxa/pxa-ssp.c
> @@ -319,6 +319,7 @@ static int pxa_ssp_set_dai_sysclk(struct snd_soc_dai *cpu_dai,
>                break;
>        case PXA_SSP_CLK_EXT:
>                priv->sysclk = freq;
> +               ssp_set_scr(&priv->dev, 4);

Shouldn't this be somehow set by set_dai_clkdiv instead?

>                sscr0 |= SSCR0_ECS;
>                break;
>        case PXA_SSP_CLK_NET:
> @@ -551,17 +552,13 @@ static int pxa_ssp_set_dai_fmt(struct snd_soc_dai *cpu_dai,
>
>        switch (fmt & SND_SOC_DAIFMT_FORMAT_MASK) {
>        case SND_SOC_DAIFMT_I2S:
> -               sscr0 |= SSCR0_MOD | SSCR0_PSP;
> +               sscr0 |= SSCR0_PSP;
>                sscr1 |= SSCR1_RWOT | SSCR1_TRAIL;
>
>                switch (fmt & SND_SOC_DAIFMT_INV_MASK) {
>                case SND_SOC_DAIFMT_NB_NF:
> -                       sspsp |= SSPSP_FSRT;
>                        break;
>                case SND_SOC_DAIFMT_NB_IF:
> -                       sspsp |= SSPSP_SFRMP | SSPSP_FSRT;
> -                       break;
> -               case SND_SOC_DAIFMT_IB_IF:
>                        sspsp |= SSPSP_SFRMP;
>                        break;

Removal of SSPSP_FSRT from NB/IB selection seems to be correct from the docs.
Can you check if IB could be properly handled by setting SCMODE(1)?

>                default:
> @@ -652,33 +649,39 @@ static int pxa_ssp_hw_params(struct snd_pcm_substream *substream,
>                break;
>        case SNDRV_PCM_FORMAT_S24_LE:
>                sscr0 |= (SSCR0_EDSS | SSCR0_DataSize(8));
> -               /* we must be in network mode (2 slots) for 24 bit stereo */

This is still dubious ...
S24_LE is 24-bit sound LSB-aligned in 32-bit frames, so DataSize
should be 32 here.

>                break;
>        case SNDRV_PCM_FORMAT_S32_LE:
>                sscr0 |= (SSCR0_EDSS | SSCR0_DataSize(16));
> -               /* we must be in network mode (2 slots) for 32 bit stereo */

How is it possible to send 64bit in one frame otherwise?

>                break;
>        }
>        ssp_write_reg(ssp, SSCR0, sscr0);
>
>        switch (priv->dai_fmt & SND_SOC_DAIFMT_FORMAT_MASK) {
>        case SND_SOC_DAIFMT_I2S:
> -               /* Cleared when the DAI format is set */
> -               sspsp = ssp_read_reg(ssp, SSPSP) | SSPSP_SFRMWDTH(width);
> +               sspsp = ssp_read_reg(ssp, SSPSP);
> +
> +               switch (priv->dai_fmt & SND_SOC_DAIFMT_FRAME_FORMAT_MASK) {
> +               case SND_SOC_DAIFMT_FF_I2S_32:
> +                       /* These values are all found out by trying and
> +                        * failing a lot. PXA's SSP is all black magic and
> +                        * does not work like described in any datasheet.
> +                        */
> +                       sspsp |= SSPSP_SFRMWDTH(32);
> +                       sspsp |= SSPSP_SFRMDLY(32 * 2);
> +                       sspsp |= SSPSP_EDMYSTOP(3);
> +                       sspsp |= SSPSP_DMYSTOP(3);
> +                       sspsp |= SSPSP_DMYSTRT(1);

Wha?! Amazing. And this really works?
How the hell can this result in 16 bits of data followed by 16 bits of
zeroes, twice :)

> +                       break;
> +               default:
> +                       /* Cleared when the DAI format is set */
> +                       sspsp |= SSPSP_SFRMWDTH(width);

Not good for DSP_A/B.

> +                       break;
> +               }
>                ssp_write_reg(ssp, SSPSP, sspsp);
> -               break;
>        default:
>                break;
>        }
>
> -       /* We always use a network mode so we always require TDM slots
> -        * - complain loudly and fail if they've not been set up yet.
> -        */
> -       if (!(ssp_read_reg(ssp, SSTSA) & 0xf)) {
> -               dev_err(&ssp->pdev->dev, "No TDM timeslot configured\n");
> -               return -EINVAL;
> -       }
> -
>        dump_registers(ssp);
>
>        return 0;
> --
> 1.6.1.3
>
>

regards
Philipp