regulator: fix: enable gpio when requested
authorMark Kettenis <kettenis@openbsd.org>
Mon, 27 Apr 2020 09:09:47 +0000 (11:09 +0200)
committerTom Rini <trini@konsulko.com>
Mon, 15 Jun 2020 20:37:01 +0000 (16:37 -0400)
The fix in commit b7adcdd073c0 has the side-effect that the regulator
will be disabled when requesting the relevant gpio in
regulator_common_ofdata_to_platdata() and enabled in
regulator_pre_probe() when the regulator was already enabled.
This leads to a short interruption in the 3.3V power to the PCIe
slot on the firefly-rk3399 which makes an ADATA SX8000NP NVMe SSD
unhappy.

Fix this by setting the GPIOD_IS_OUT_ACTIVE flag again when the
'regulator-boot-on' property is set, but check for this property
explicitly instead of relying on the "boot_on" member of
the uclass platdata.

Signed-off-by: Mark Kettenis <kettenis@openbsd.org>
Tested-by: Patrice Chotard <patrice.chotard@st.com>
drivers/power/regulator/regulator-uclass.c
drivers/power/regulator/regulator_common.c

index 7dcb3171928c4d6b515f005e38f029848a73ac83..da679a572c23b0d1002c6aa87c371595dbe0e7ab 100644 (file)
@@ -466,9 +466,6 @@ static int regulator_pre_probe(struct udevice *dev)
            (uc_pdata->min_uA == uc_pdata->max_uA))
                uc_pdata->flags |= REGULATOR_FLAG_AUTOSET_UA;
 
-       if (uc_pdata->boot_on)
-               regulator_set_enable(dev, uc_pdata->boot_on);
-
        return 0;
 }
 
index 16d9412cd52d1f78c202404161815302be6e8e8d..4cfcc31298eadb03aaa41173b159e131c9ce391f 100644 (file)
@@ -19,6 +19,8 @@ int regulator_common_ofdata_to_platdata(struct udevice *dev,
 
        if (!dev_read_bool(dev, "enable-active-high"))
                flags |= GPIOD_ACTIVE_LOW;
+       if (dev_read_bool(dev, "regulator-boot-on"))
+               flags |= GPIOD_IS_OUT_ACTIVE;
 
        /* Get optional enable GPIO desc */
        gpio = &dev_pdata->gpio;