gpio: gpio-dwapb: add optional reset

Message ID 20171010212617.3397-1-atull@kernel.org
State New
Headers show
Series
  • gpio: gpio-dwapb: add optional reset
Related show

Commit Message

Alan Tull Oct. 10, 2017, 9:26 p.m.
Some platforms require reset to be released to allow register
access.

Signed-off-by: Alan Tull <atull@kernel.org>
---
 drivers/gpio/gpio-dwapb.c | 14 ++++++++++++++
 1 file changed, 14 insertions(+)

Comments

Linus Walleij Oct. 11, 2017, 8:31 a.m. | #1
On Tue, Oct 10, 2017 at 11:26 PM, Alan Tull <atull@kernel.org> wrote:

> Some platforms require reset to be released to allow register
> access.
>
> Signed-off-by: Alan Tull <atull@kernel.org>

Fair enough.

(...)
> +       rst = devm_reset_control_get_optional_exclusive(dev, NULL);
> +       if (IS_ERR(rst)) {
> +               if (PTR_ERR(rst) == -EPROBE_DEFER)
> +                       return PTR_ERR(rst);
> +       } else {
> +               reset_control_deassert(rst);
> +               gpio->rst = rst;
> +       }

I do not see why any error other than -EPROBE_DEFER
should be ignored?

I guess the _optional API returns NULL if there is no
reset line so it should be fine to just return the error on
any error.

> +       if (gpio->rst)
> +               reset_control_assert(gpio->rst);

Is this the right way to handle an optional reset line?

Yours,
Linus Walleij
--
To unsubscribe from this list: send the line "unsubscribe linux-gpio" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Philipp Zabel Oct. 11, 2017, 10:56 a.m. | #2
Hi Alan, Linus,

On Wed, 2017-10-11 at 10:31 +0200, Linus Walleij wrote:
> On Tue, Oct 10, 2017 at 11:26 PM, Alan Tull <atull@kernel.org> wrote:
> 
> > Some platforms require reset to be released to allow register
> > access.
> > 
> > Signed-off-by: Alan Tull <atull@kernel.org>
> 
> Fair enough.
> 
> (...)
> > +       rst = devm_reset_control_get_optional_exclusive(dev, NULL);

The way this reset control is used, it looks like you could use _shared
instead of _exclusive here. This relaxes the guarantees made by the API
a bit and may allow this driver to work with more reset controllers.

> > +       if (IS_ERR(rst)) {
> > +               if (PTR_ERR(rst) == -EPROBE_DEFER)
> > +                       return PTR_ERR(rst);

The _optional variant of reset_control_get returns NULL if no reset is
specified in the device tree. If an error value is returned, it is
always an actual error (invalid device tree contents, reset is specified
in the device tree but the driver returns an error, etc.).
This should just be:

	if (IS_ERR(rst))
		return PTR_ERR(rst);

> > +       } else {
> > +               reset_control_deassert(rst);
> > +               gpio->rst = rst;

And this should be made unconditional. reset_control_deassert just
ignores rst == NULL.

> > +       }
> 
> I do not see why any error other than -EPROBE_DEFER
> should be ignored?
> 
> I guess the _optional API returns NULL if there is no
> reset line so it should be fine to just return the error on
> any error.

Correct. The _optional API together with NULL reset control handles
allows to simplify handling of optional resets in the consumer drivers.

> > +       if (gpio->rst)
> > +               reset_control_assert(gpio->rst);
> 
> Is this the right way to handle an optional reset line?

Just as the deassert above, this should be made unconditional.

regards
Philipp
--
To unsubscribe from this list: send the line "unsubscribe linux-gpio" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Alan Tull Oct. 11, 2017, 4:05 p.m. | #3
On Wed, Oct 11, 2017 at 5:56 AM, Philipp Zabel <p.zabel@pengutronix.de> wrote:

Hi Linus, Phillipp,

Thanks for the review.

> Hi Alan, Linus,
>
> On Wed, 2017-10-11 at 10:31 +0200, Linus Walleij wrote:
>> On Tue, Oct 10, 2017 at 11:26 PM, Alan Tull <atull@kernel.org> wrote:
>>
>> > Some platforms require reset to be released to allow register
>> > access.
>> >
>> > Signed-off-by: Alan Tull <atull@kernel.org>
>>
>> Fair enough.
>>
>> (...)
>> > +       rst = devm_reset_control_get_optional_exclusive(dev, NULL);
>
> The way this reset control is used, it looks like you could use _shared
> instead of _exclusive here. This relaxes the guarantees made by the API
> a bit and may allow this driver to work with more reset controllers.

OK, will use devm_reset_control_get_optional_shared().

>
>> > +       if (IS_ERR(rst)) {
>> > +               if (PTR_ERR(rst) == -EPROBE_DEFER)
>> > +                       return PTR_ERR(rst);
>
> The _optional variant of reset_control_get returns NULL if no reset is
> specified in the device tree. If an error value is returned, it is
> always an actual error (invalid device tree contents, reset is specified
> in the device tree but the driver returns an error, etc.).
> This should just be:
>
>         if (IS_ERR(rst))
>                 return PTR_ERR(rst);

That's great!

>
>> > +       } else {
>> > +               reset_control_deassert(rst);
>> > +               gpio->rst = rst;
>
> And this should be made unconditional. reset_control_deassert just
> ignores rst == NULL.

Nice.

>
>> > +       }
>>
>> I do not see why any error other than -EPROBE_DEFER
>> should be ignored?
>>
>> I guess the _optional API returns NULL if there is no
>> reset line so it should be fine to just return the error on
>> any error.
>
> Correct. The _optional API together with NULL reset control handles
> allows to simplify handling of optional resets in the consumer drivers.
>
>> > +       if (gpio->rst)
>> > +               reset_control_assert(gpio->rst);
>>
>> Is this the right way to handle an optional reset line?
>
> Just as the deassert above, this should be made unconditional.

I've made the requested changes which shrinks the patch to be even smaller :)

Will send up v2 soon.

Thanks,
Alan

>
> regards
> Philipp
--
To unsubscribe from this list: send the line "unsubscribe linux-gpio" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

diff --git a/drivers/gpio/gpio-dwapb.c b/drivers/gpio/gpio-dwapb.c
index 5cdb7bc..f7ac625 100644
--- a/drivers/gpio/gpio-dwapb.c
+++ b/drivers/gpio/gpio-dwapb.c
@@ -25,6 +25,7 @@ 
 #include <linux/of_irq.h>
 #include <linux/platform_device.h>
 #include <linux/property.h>
+#include <linux/reset.h>
 #include <linux/spinlock.h>
 #include <linux/platform_data/gpio-dwapb.h>
 #include <linux/slab.h>
@@ -98,6 +99,7 @@  struct dwapb_gpio {
 	unsigned int		nr_ports;
 	struct irq_domain	*domain;
 	unsigned int		flags;
+	struct reset_control	*rst;
 };
 
 static inline u32 gpio_reg_v2_convert(unsigned int offset)
@@ -612,6 +614,7 @@  static int dwapb_gpio_probe(struct platform_device *pdev)
 	int err;
 	struct device *dev = &pdev->dev;
 	struct dwapb_platform_data *pdata = dev_get_platdata(dev);
+	struct reset_control *rst;
 
 	if (!pdata) {
 		pdata = dwapb_gpio_get_pdata(dev);
@@ -629,6 +632,15 @@  static int dwapb_gpio_probe(struct platform_device *pdev)
 	gpio->dev = &pdev->dev;
 	gpio->nr_ports = pdata->nports;
 
+	rst = devm_reset_control_get_optional_exclusive(dev, NULL);
+	if (IS_ERR(rst)) {
+		if (PTR_ERR(rst) == -EPROBE_DEFER)
+			return PTR_ERR(rst);
+	} else {
+		reset_control_deassert(rst);
+		gpio->rst = rst;
+	}
+
 	gpio->ports = devm_kcalloc(&pdev->dev, gpio->nr_ports,
 				   sizeof(*gpio->ports), GFP_KERNEL);
 	if (!gpio->ports)
@@ -680,6 +692,8 @@  static int dwapb_gpio_remove(struct platform_device *pdev)
 
 	dwapb_gpio_unregister(gpio);
 	dwapb_irq_teardown(gpio);
+	if (gpio->rst)
+		reset_control_assert(gpio->rst);
 
 	return 0;
 }