Re: [PATCH V6 5/6] drm: bridge: samsung-dsim: Dynamically configure DPHY timing

[Adam Ford]

On Wed, May 17, 2023 at 6:28 AM Jagan Teki <jagan@xxxxxxxxxxxxxxxxxxxx> wrote:
>
> On Tue, May 16, 2023 at 5:27 AM Adam Ford <aford173@xxxxxxxxx> wrote:
> >
> > The DPHY timings are currently hard coded. Since the input
> > clock can be variable, the phy timings need to be variable
> > too.  To facilitate this, we need to cache the hs_clock
> > based on what is generated from the PLL.
> >
> > The phy_mipi_dphy_get_default_config_for_hsclk function
> > configures the DPHY timings in pico-seconds, and a small macro
> > converts those timings into clock cycles based on the hs_clk.
> >
> > Signed-off-by: Adam Ford <aford173@xxxxxxxxx>
> > Signed-off-by: Lucas Stach <l.stach@xxxxxxxxxxxxxx>
> > Tested-by: Chen-Yu Tsai <wenst@xxxxxxxxxxxx>
> > Tested-by: Frieder Schrempf <frieder.schrempf@xxxxxxxxxx>
> > Reviewed-by: Frieder Schrempf <frieder.schrempf@xxxxxxxxxx>
> > Tested-by: Michael Walle <michael@xxxxxxxx>
> > ---
> >  drivers/gpu/drm/bridge/samsung-dsim.c | 57 +++++++++++++++++++++++----
> >  include/drm/bridge/samsung-dsim.h     |  1 +
> >  2 files changed, 51 insertions(+), 7 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/bridge/samsung-dsim.c b/drivers/gpu/drm/bridge/samsung-dsim.c
> > index 08266303c261..3944b7cfbbdf 100644
> > --- a/drivers/gpu/drm/bridge/samsung-dsim.c
> > +++ b/drivers/gpu/drm/bridge/samsung-dsim.c
> > @@ -218,6 +218,8 @@
> >
> >  #define OLD_SCLK_MIPI_CLK_NAME         "pll_clk"
> >
> > +#define PS_TO_CYCLE(ps, hz) DIV64_U64_ROUND_CLOSEST(((ps) * (hz)), 1000000000000ULL)
> > +
> >  static const char *const clk_names[5] = {
> >         "bus_clk",
> >         "sclk_mipi",
> > @@ -651,6 +653,8 @@ static unsigned long samsung_dsim_set_pll(struct samsung_dsim *dsi,
> >                 reg = samsung_dsim_read(dsi, DSIM_STATUS_REG);
> >         } while ((reg & DSIM_PLL_STABLE) == 0);
> >
> > +       dsi->hs_clock = fout;
> > +
> >         return fout;
> >  }
> >
> > @@ -698,13 +702,46 @@ static void samsung_dsim_set_phy_ctrl(struct samsung_dsim *dsi)
> >         const struct samsung_dsim_driver_data *driver_data = dsi->driver_data;
> >         const unsigned int *reg_values = driver_data->reg_values;
> >         u32 reg;
> > +       struct phy_configure_opts_mipi_dphy cfg;
> > +       int clk_prepare, lpx, clk_zero, clk_post, clk_trail;
> > +       int hs_exit, hs_prepare, hs_zero, hs_trail;
> > +       unsigned long long byte_clock = dsi->hs_clock / 8;
> >
> >         if (driver_data->has_freqband)
> >                 return;
> >
> > +       phy_mipi_dphy_get_default_config_for_hsclk(dsi->hs_clock,
> > +                                                  dsi->lanes, &cfg);
> > +
> > +       /*
> > +        * TODO:
> > +        * The tech reference manual for i.MX8M Mini/Nano/Plus
>
> Does it mean, Applications Processor Reference Manual? better add it
> clear reference.

I can do that.

>
> > +        * doesn't state what the definition of the PHYTIMING
> > +        * bits are beyond their address and bit position.
> > +        * After reviewing NXP's downstream code, it appears
> > +        * that the various PHYTIMING registers take the number
> > +        * of cycles and use various dividers on them.  This
> > +        * calculation does not result in an exact match to the
> > +        * downstream code, but it is very close, and it appears
> > +        * to sync at a variety of resolutions. If someone
> > +        * can get a more accurate mathematical equation needed
> > +        * for these registers, this should be updated.
> > +        */
> > +
> > +       lpx = PS_TO_CYCLE(cfg.lpx, byte_clock);
> > +       hs_exit = PS_TO_CYCLE(cfg.hs_exit, byte_clock);
> > +       clk_prepare = PS_TO_CYCLE(cfg.clk_prepare, byte_clock);
> > +       clk_zero = PS_TO_CYCLE(cfg.clk_zero, byte_clock);
> > +       clk_post = PS_TO_CYCLE(cfg.clk_post, byte_clock);
> > +       clk_trail = PS_TO_CYCLE(cfg.clk_trail, byte_clock);
> > +       hs_prepare = PS_TO_CYCLE(cfg.hs_prepare, byte_clock);
> > +       hs_zero = PS_TO_CYCLE(cfg.hs_zero, byte_clock);
> > +       hs_trail = PS_TO_CYCLE(cfg.hs_trail, byte_clock);
>
> I think we can do some kind of negotiation has done similar in bsp by
> taking inputs from bit_clk  and PLL_1432X table. Did you try this? we
> thought this approach while writing dsim to support dynamic dphy.

I originally attempted to implement the lookup table that was used in
the downstream NXP kernel, but I was told to not use a lookup table
but to calculate them directly instead.  I reached out to my NXP rep
and I was told that they could not divulge the contents of these
registers since the DSI driver was under a license from Samsung and
that information was not available outside of an NDA.

When I did the testing for this, I tested a variety of resolutions and
refresh rates and compared the values output from here to those
generated by the NXP lookup table, and they are very close.  From what
I could tell, the variance didn't appear to manifest itself on the
monitors that I tried.  I tried to explain this in the TODO message,
but maybe it wasn't clear.

Marek S tested this on Exynos and he didn't report any regressions.

adam
>
> Thanks,
> Jagan.