[PATCH] i2c: rcar: fix reset handling for Gen5

Missing reset_control_status() support is not Gen5 specific. It depends
on the firmware used, if any. Refactor the code to handle missing
reset_control_status() more generically.

Fixes: 87e713f20048 ("i2c: rcar: add R-Car Gen5 support")
Suggested-by: Geert Uytterhoeven <[email protected]>
Signed-off-by: Wolfram Sang <[email protected]>
---

Based on this discussion[1]. Tested on a Ironhide board (R-Car X5H) and
a SparrowHawk (R-Car V4H) for regression testing.

[1] https://lore.kernel.org/r/CAMuHMdVAgjguscFMKd7u1XPuZEkB7mH9-G-Y8nYfXqbzJ2hH9Q@mail.gmail.com

 drivers/i2c/busses/i2c-rcar.c | 27 ++++++++++++---------------
 1 file changed, 12 insertions(+), 15 deletions(-)

diff --git a/drivers/i2c/busses/i2c-rcar.c b/drivers/i2c/busses/i2c-rcar.c
index 46508176712c..755064804cb2 100644
--- a/drivers/i2c/busses/i2c-rcar.c
+++ b/drivers/i2c/busses/i2c-rcar.c
@@ -123,12 +123,13 @@
 #define ID_NACK			BIT(4)
 #define ID_EPROTO		BIT(5)
 /* persistent flags */
+#define ID_P_NO_RST_STAT	BIT(26)
 #define ID_P_FMPLUS		BIT(27)
 #define ID_P_NOT_ATOMIC		BIT(28)
 #define ID_P_HOST_NOTIFY	BIT(29)
 #define ID_P_NO_RXDMA		BIT(30) /* HW forbids RXDMA sometimes */
 #define ID_P_PM_BLOCKED		BIT(31)
-#define ID_P_MASK		GENMASK(31, 27)
+#define ID_P_MASK		GENMASK(31, 26)
 
 #define ID_SLAVE_NACK		BIT(0)
 
@@ -901,12 +902,11 @@ static int rcar_i2c_do_reset(struct rcar_i2c_priv *priv)
 	if (ret)
 		return ret;
 
-	/* SCMI based resets don't need to poll for success */
-	if (priv->devtype < I2C_RCAR_GEN5)
-		return read_poll_timeout_atomic(reset_control_status, ret, ret == 0,
-						1, 100, false, priv->rstc);
+	if (priv->flags & ID_P_NO_RST_STAT)
+		return 0;
 
-	return 0;
+	return read_poll_timeout_atomic(reset_control_status, ret, ret == 0,
+					1, 100, false, priv->rstc);
 }
 
 static int rcar_i2c_master_xfer(struct i2c_adapter *adap,
@@ -1200,15 +1200,12 @@ static int rcar_i2c_probe(struct platform_device *pdev)
 			goto out_pm_put;
 		}
 
-		/*
-		 * Gen5+ uses SCMI based reset which cannot report status.
-		 * Firmware has to ensure proper reset
-		 */
-		if (priv->devtype < I2C_RCAR_GEN5) {
-			ret = reset_control_status(priv->rstc);
-			if (ret < 0)
-				goto out_pm_put;
-		}
+		ret = reset_control_status(priv->rstc);
+		/* Some SCMI firmware does not support reading reset status */
+		if (ret == -ENOTSUPP)
+			priv->flags |= ID_P_NO_RST_STAT;
+		else if (ret < 0)
+			goto out_pm_put;
 
 		/* hard reset disturbs HostNotify local target, so disable it */
 		priv->flags &= ~ID_P_HOST_NOTIFY;
-- 
2.47.3