[PATCH v2] mmc: mmci.c: Defer probe() in case of yet uninitialized GPIOs

[PATCH v2] mmc: mmci.c: Defer probe() in case of yet uninitialized GPIOs

[Roland Stigge]

If the GPIOs used by the MMCI driver are not registered yet when the driver is
probe()d, they can't be used. This happens if the mmci driver is probed before
the respective GPIO controller (e.g. on the LPC32xx EA3250 board, the PCA9532
GPIO controller would be initialized via DT after mmci). Therefore, we defer
mmci in this case.

Signed-off-by: Roland Stigge <stigge@antcom.de>

---
 drivers/mmc/host/mmci.c |   30 +++++++++++++++++++++++-------
 1 file changed, 23 insertions(+), 7 deletions(-)

--- linux-2.6.orig/drivers/mmc/host/mmci.c
+++ linux-2.6/drivers/mmc/host/mmci.c
@@ -1215,13 +1215,21 @@ static void mmci_dt_populate_generic_pda
 {
 	int bus_width = 0;
 
-	pdata->gpio_wp = of_get_named_gpio(np, "wp-gpios", 0);
-	if (!pdata->gpio_wp)
-		pdata->gpio_wp = -1;
-
-	pdata->gpio_cd = of_get_named_gpio(np, "cd-gpios", 0);
-	if (!pdata->gpio_cd)
-		pdata->gpio_cd = -1;
+	if (of_get_property(np, "wp-gpios", NULL)) {
+		pdata->gpio_wp = of_get_named_gpio(np, "wp-gpios", 0);
+		if (pdata->gpio_wp == -ENODEV)
+			pdata->gpio_wp = -EPROBE_DEFER;
+	} else {
+		pdata->gpio_wp = -ENODEV;
+	}
+
+	if (of_get_property(np, "cd-gpios", NULL)) {
+		pdata->gpio_cd = of_get_named_gpio(np, "cd-gpios", 0);
+		if (pdata->gpio_cd == -ENODEV)
+			pdata->gpio_cd = -EPROBE_DEFER;
+	} else {
+		pdata->gpio_cd = -ENODEV;
+	}
 
 	if (of_get_property(np, "cd-inverted", NULL))
 		pdata->cd_invert = true;
@@ -1424,6 +1432,10 @@ static int __devinit mmci_probe(struct a
 	writel(0, host->base + MMCIMASK1);
 	writel(0xfff, host->base + MMCICLEAR);
 
+	if (plat->gpio_cd == -EPROBE_DEFER) {
+		ret = -EPROBE_DEFER;
+		goto err_gpio_cd;
+	}
 	if (gpio_is_valid(plat->gpio_cd)) {
 		ret = gpio_request(plat->gpio_cd, DRIVER_NAME " (cd)");
 		if (ret == 0)
@@ -1447,6 +1459,10 @@ static int __devinit mmci_probe(struct a
 		if (ret >= 0)
 			host->gpio_cd_irq = gpio_to_irq(plat->gpio_cd);
 	}
+	if (plat->gpio_wp == -EPROBE_DEFER) {
+		ret = -EPROBE_DEFER;
+		goto err_gpio_wp;
+	}
 	if (gpio_is_valid(plat->gpio_wp)) {
 		ret = gpio_request(plat->gpio_wp, DRIVER_NAME " (wp)");
 		if (ret == 0)

[PATCH v2] mmc: mmci.c: Defer probe() in case of yet uninitialized GPIOs

[Arnd Bergmann]

On Saturday 16 June 2012, Roland Stigge wrote:
> +       if (of_get_property(np, "wp-gpios", NULL)) {
> +               pdata->gpio_wp = of_get_named_gpio(np, "wp-gpios", 0);
> +               if (pdata->gpio_wp == -ENODEV)
> +                       pdata->gpio_wp = -EPROBE_DEFER;
> +       } else {
> +               pdata->gpio_wp = -ENODEV;
> +       }

This is still very ugly and there is a certain danfer that we might
have to duplicate this across more drivers. Can't we make
of_get_named_gpio_flags return -EPROBE_DEFER in cases where we can't
find the GPIO number yet but still have a chance of returning it at
a later point?

	Arnd

[PATCH v2] mmc: mmci.c: Defer probe() in case of yet uninitialized GPIOs

[Roland Stigge]

On 16/06/12 19:23, Arnd Bergmann wrote:
> On Saturday 16 June 2012, Roland Stigge wrote:
>> +       if (of_get_property(np, "wp-gpios", NULL)) {
>> +               pdata->gpio_wp = of_get_named_gpio(np, "wp-gpios", 0);
>> +               if (pdata->gpio_wp == -ENODEV)
>> +                       pdata->gpio_wp = -EPROBE_DEFER;
>> +       } else {
>> +               pdata->gpio_wp = -ENODEV;
>> +       }
> 
> This is still very ugly and there is a certain danfer that we might
> have to duplicate this across more drivers. Can't we make
> of_get_named_gpio_flags return -EPROBE_DEFER in cases where we can't
> find the GPIO number yet but still have a chance of returning it at
> a later point?

OK, will post a new patch.

Roland