[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAMuHMdWO2KVpg50Zxak6e_7-5c5Xg6O2PJkPkTfrR=treOtUig@mail.gmail.com>
Date: Tue, 19 Sep 2023 11:14:25 +0200
From: Geert Uytterhoeven <geert@...ux-m68k.org>
To: Wolfram Sang <wsa+renesas@...g-engineering.com>
Cc: linux-renesas-soc@...r.kernel.org,
Andi Shyti <andi.shyti@...nel.org>, linux-i2c@...r.kernel.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH RFT 2/2] i2c: rcar: improve accuracy for R-Car Gen3+
Hi Wolfram,
On Wed, Sep 13, 2023 at 11:38 AM Wolfram Sang
<wsa+renesas@...g-engineering.com> wrote:
> With some new registers, SCL can be calculated to be closer to the
> desired rate. Apply the new formula for R-Car Gen3 device types.
>
> Signed-off-by: Wolfram Sang <wsa+renesas@...g-engineering.com>
Thanks for your patch!
> --- a/drivers/i2c/busses/i2c-rcar.c
> +++ b/drivers/i2c/busses/i2c-rcar.c
> @@ -84,11 +88,25 @@
> #define RMDMAE BIT(1) /* DMA Master Received Enable */
> #define TMDMAE BIT(0) /* DMA Master Transmitted Enable */
>
> +/* ICCCR2 */
> +#define CDFD BIT(2) /* CDF Disable */
> +#define HLSE BIT(1) /* HIGH/LOW Separate Control Enable */
> +#define SME BIT(0) /* SCL Mask Enable */
> +
> /* ICFBSCR */
> #define TCYC17 0x0f /* 17*Tcyc delay 1st bit between SDA and SCL */
>
> #define RCAR_MIN_DMA_LEN 8
>
> +/* SCL low/high ratio 5:4 to meet all I2C timing specs (incl safety margin) */
> +#define RCAR_SCLD_RATIO 5
> +#define RCAR_SCHD_RATIO 4
> +/*
> + * SMD should be smaller than SCLD/SCHD and is always around 20 in the docs.
> + * Thus, we simply use 20 which works for low and high speeds.
> +*/
> +#define RCAR_DEFAULT_SMD 20
> +
> #define RCAR_BUS_PHASE_START (MDBS | MIE | ESG)
> #define RCAR_BUS_PHASE_DATA (MDBS | MIE)
> #define RCAR_BUS_PHASE_STOP (MDBS | MIE | FSB)
> @@ -128,6 +146,7 @@ struct rcar_i2c_priv {
>
> int pos;
> u32 icccr;
> + u32 scl_gran;
You may want to store just u16 schd and scld instead, so you don't
have to calculate them over and over again in rcar_i2c_init().
They are calculated in rcar_i2c_clock_calculate() (called from .probe())
anyway to validate ranges.
That would also avoid the need to come up with a better name for
scl_gran ;-)
> u8 recovery_icmcr; /* protected by adapter lock */
> enum rcar_i2c_type devtype;
> struct i2c_client *slave;
> @@ -216,11 +235,16 @@ static void rcar_i2c_init(struct rcar_i2c_priv *priv)
> rcar_i2c_write(priv, ICMCR, MDBS);
> rcar_i2c_write(priv, ICMSR, 0);
> /* start clock */
> - rcar_i2c_write(priv, ICCCR, priv->icccr);
> -
> - if (priv->devtype == I2C_RCAR_GEN3)
> + if (priv->devtype < I2C_RCAR_GEN3) {
> + rcar_i2c_write(priv, ICCCR, priv->icccr);
> + } else {
> + rcar_i2c_write(priv, ICCCR2, CDFD | HLSE | SME);
> + rcar_i2c_write(priv, ICCCR, priv->icccr);
> + rcar_i2c_write(priv, ICMPR, RCAR_DEFAULT_SMD);
> + rcar_i2c_write(priv, ICHPR, RCAR_SCHD_RATIO * priv->scl_gran);
> + rcar_i2c_write(priv, ICLPR, RCAR_SCLD_RATIO * priv->scl_gran);
> rcar_i2c_write(priv, ICFBSCR, TCYC17);
> -
> + }
> }
>
> static int rcar_i2c_bus_barrier(struct rcar_i2c_priv *priv)
> @@ -301,24 +316,57 @@ static int rcar_i2c_clock_calculate(struct rcar_i2c_priv *priv)
> round = DIV_ROUND_CLOSEST(ick, 1000000);
> round = DIV_ROUND_CLOSEST(round * sum, 1000);
>
> - /*
> - * SCL = ick / (20 + 8 * SCGD + F[(ticf + tr + intd) * ick])
> - * 20 + 8 * SCGD + F[...] = ick / SCL
> - * SCGD = ((ick / SCL) - 20 - F[...]) / 8
> - * Result (= SCL) should be less than bus_speed for hardware safety
> - */
> - scgd = DIV_ROUND_UP(ick, t.bus_freq_hz ?: 1);
> - scgd = DIV_ROUND_UP(scgd - 20 - round, 8);
> - scl = ick / (20 + 8 * scgd + round);
> + if (priv->devtype < I2C_RCAR_GEN3) {
> + u32 scgd;
> + /*
> + * SCL = ick / (20 + 8 * SCGD + F[(ticf + tr + intd) * ick])
> + * 20 + 8 * SCGD + F[...] = ick / SCL
> + * SCGD = ((ick / SCL) - 20 - F[...]) / 8
> + * Result (= SCL) should be less than bus_speed for hardware safety
> + */
> + scgd = DIV_ROUND_UP(ick, t.bus_freq_hz ?: 1);
> + scgd = DIV_ROUND_UP(scgd - 20 - round, 8);
> + scl = ick / (20 + 8 * scgd + round);
>
> - if (scgd > 0x3f)
> - goto err_no_val;
> + if (scgd > 0x3f)
> + goto err_no_val;
>
> - dev_dbg(dev, "clk %u/%u(%lu), round %u, CDF: %u, SCGD: %u\n",
> - scl, t.bus_freq_hz, rate, round, cdf, scgd);
> + dev_dbg(dev, "clk %u/%u(%lu), round %u, CDF: %u, SCGD: %u\n",
> + scl, t.bus_freq_hz, rate, round, cdf, scgd);
>
> - /* keep icccr value */
> - priv->icccr = scgd << cdf_width | cdf;
> + priv->icccr = scgd << cdf_width | cdf;
> + } else {
> + u32 x, sum_ratio = RCAR_SCHD_RATIO + RCAR_SCLD_RATIO;
> + /*
> + * SCLD/SCHD ratio and SMD default value are explained above
> + * where they are defined. With these definitions, we can compute
> + * x as a base value for the SCLD/SCHD ratio:
> + *
> + * SCL = clkp / (8 + 2 * SMD + SCLD + SCHD + F[(ticf + tr + intd) * clkp])
> + * SCL = clkp / (8 + 2 * RCAR_DEFAULT_SMD + RCAR_SCLD_RATIO * x
> + * + RCAR_SCHD_RATIO * x + F[...])
> + *
> + * with: sum_ratio = RCAR_SCLD_RATIO + RCAR_SCHD_RATIO
> + * and: smd = 2 * RCAR_DEFAULT_SMD
> + *
> + * SCL = clkp / (8 + smd + sum_ratio * x + F[...])
> + * 8 + smd + sum_ratio * x + F[...] = SCL / clkp
> + * x = ((SCL / clkp) - 8 - smd - F[...]) / sum_ratio
> + */
> + x = DIV_ROUND_UP(rate, t.bus_freq_hz ?: 1);
> + x = DIV_ROUND_UP(x - 8 - 2 * RCAR_DEFAULT_SMD - round, sum_ratio);
> + scl = rate / (8 + 2 * RCAR_DEFAULT_SMD + sum_ratio * x + round);
> +
> + /* Bail out if values don't fit into 16 bit or SMD became too large */
> + if (x * RCAR_SCLD_RATIO > 0xffff || RCAR_DEFAULT_SMD > x * RCAR_SCHD_RATIO)
The second part of the check looks wrong to me, as it would reject
all the recommended register values for SMD and SCHD in the docs .
What does "SMD became too large" mean here?
> + goto err_no_val;
> +
> + dev_dbg(dev, "clk %u/%u(%lu), round %u, CDF: %u SCL gran %u\n",
> + scl, t.bus_freq_hz, rate, round, cdf, x);
> +
> + priv->icccr = cdf;
> + priv->scl_gran = x;
> + }
>
> return 0;
The rest LGTM.
BTW, Note 2 in the docs for the Clock Control Register 2 (ICCCR2) seems
to be incorrect:
SCLD: I2C SCL low clock period (set by the SCL HIGH control register)
^^^ ^^^^
SCHD: I2C SCL high clock period (set by the SCL LOW control register)
^^^^ ^^^
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@...ux-m68k.org
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
Powered by blists - more mailing lists