mfd: syscon: enable specified clocks on syscon_base_lookup_by_phandle

Since commit b36b7b72 ("mfd: syscon: clock peripheral if specified in
device tree"), we now clock syscons during access if the device tree nodes
indicate a clocks property. We haven't been doing this for
syscon_base_lookup_by_phandle though, because we did this as part of the
regmap access functions, but that function returns a direct pointer to
the MMIO region.

The best way forward is probably dropping the syscon_base API altogether
and change users to the regmap API instead, but for now, make the behavior
consistent by enabling it permanently.

This makes use safe from breakage that results from upstream device trees moving
a clock from the consumer to the syscon provider (like in [1])

I've reviewed the current driver code users of syscon_base_lookup_by_phandle
and all upstream device trees that match against the drivers lack
a clocks property, so this shouldn't alter behavior (for now).

[1]: c9322d4fe ("net: designware: eqos: stm32: drop no longer needed syscfg-clk")

Signed-off-by: Ahmad Fatoum <ahmad@a3f.at>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

1 files changed, 10 insertions, 0 deletions

diff --git a/drivers/mfd/syscon.c b/drivers/mfd/syscon.c
index f1e6559d71..ab458428a6 100644
--- a/drivers/mfd/syscon.c
+++ b/drivers/mfd/syscon.c
@@ -105,10 +105,20 @@ static struct syscon *node_to_syscon(struct device_node *np)
 static void __iomem *syscon_node_to_base(struct device_node *np)
 {
 	struct syscon *syscon = node_to_syscon(np);
+	struct clk *clk;
 
 	if (IS_ERR(syscon))
 		return ERR_CAST(syscon);
 
+	/* Returning the direct pointer here side steps the regmap
+	 * and any specified clock wouldn't be enabled on access.
+	 * Deal with this by enabling the clock permanently if any
+	 * syscon_node_to_base users exist.
+	 */
+	clk = of_clk_get(np, 0);
+	if (!IS_ERR(clk))
+		clk_enable(clk);
+
 	return syscon->base;
 }