I recently ran into a problem with gpio-generic driver on mxs platform.
It turns out it's the exactly same problem that Lothar had run into
on imx/mxc platform, which was fixed up in gpio-mxc driver as below.
commit fb1492186276ba52d99b58121b8a9a87f20cc9f3
Author: Lothar Waßmann <[email protected]>
Date: Thu Jul 7 14:50:16 2011 +0200
gpio/mxc: add missing initialization of basic_mmio_gpio shadow variables
The bgpio_init() function does not initialise the shadow register for
the GPIO direction register. Thus, when configuring the first GPIO with
gpio_set_direction() all other GPIOs of the same bank will be
configured as inputs. Since the bgpio layer cannot know whether the
register is readable, the initialisation should be done by the caller
of bgpio_init().
Also, the 'data' shadow variable that is used inside basic_mmio_gpio
to cache the current value of the GPIO_DR register is initialised from
the GPIO_PSR register within bgpio_init(). Thus when setting the
output value of a certain GPIO, the other GPIO outputs of the same
bank will be set or cleared depending on the pin state of the GPIO
inputs during bgpio_init().
I think it's a problem of gpio-generic rather than platform specific
driver, so I'm coming up this series to have the problem fixed in
gpio-generic driver itself and then removes the fixup in gpio-mxc.
Thus, other platform gpio drivers based on gpio-generic do not have to
suffer from the same problem and fix it up in individually over and
over again.
Per suggestion from Lothar, I'm copying the maintainers of those
platform gpio drivers using gpio-generic to inform the default behavior
changing on gpio-generic. Please let me know if anyone gets impacted
by the changes.
Regards,
Shawn
Shawn Guo (2):
gpio/generic: initialize basic_mmio_gpio shadow variables properly
gpio/mxc: remove basic_mmio_gpio shadow fixup
drivers/gpio/gpio-generic.c | 4 ++++
drivers/gpio/gpio-mxc.c | 2 --
2 files changed, 4 insertions(+), 2 deletions(-)
--
1.7.5.4
It fixes the issue in gpio-generic that commit fb14921 (gpio/mxc: add
missing initialization of basic_mmio_gpio shadow variables) manged to
fix in gpio-mxc driver, so that other platform specific drivers do not
suffer from the same problem over and over again.
Signed-off-by: Shawn Guo <[email protected]>
---
drivers/gpio/gpio-generic.c | 4 ++++
1 files changed, 4 insertions(+), 0 deletions(-)
diff --git a/drivers/gpio/gpio-generic.c b/drivers/gpio/gpio-generic.c
index e38dd0c..cfc9439 100644
--- a/drivers/gpio/gpio-generic.c
+++ b/drivers/gpio/gpio-generic.c
@@ -394,6 +394,10 @@ int bgpio_init(struct bgpio_chip *bgc, struct device *dev,
return ret;
bgc->data = bgc->read_reg(bgc->reg_dat);
+ if (bgc->gc.set == bgpio_set_set)
+ bgc->data = bgc->read_reg(bgc->reg_set);
+ if (bgc->reg_dir)
+ bgc->dir = bgc->read_reg(bgc->reg_dir);
return ret;
}
--
1.7.5.4
With commit 7cc9c7c (gpio/generic: initialize basic_mmio_gpio shadow
variables properly) fixing the basic_mmio_gpio shadow variables in
gpio-generic, the fixup in gpio-mxc can be removed now.
Signed-off-by: Shawn Guo <[email protected]>
---
drivers/gpio/gpio-mxc.c | 2 --
1 files changed, 0 insertions(+), 2 deletions(-)
diff --git a/drivers/gpio/gpio-mxc.c b/drivers/gpio/gpio-mxc.c
index e791476..dbaac01 100644
--- a/drivers/gpio/gpio-mxc.c
+++ b/drivers/gpio/gpio-mxc.c
@@ -423,8 +423,6 @@ static int __devinit mxc_gpio_probe(struct platform_device *pdev)
port->bgc.gc.to_irq = mxc_gpio_to_irq;
port->bgc.gc.base = pdev->id * 32;
- port->bgc.dir = port->bgc.read_reg(port->bgc.reg_dir);
- port->bgc.data = port->bgc.read_reg(port->bgc.reg_set);
err = gpiochip_add(&port->bgc.gc);
if (err)
--
1.7.5.4
On Thursday, May 17, 2012 12:35 AM, Shawn Guo wrote:
> It fixes the issue in gpio-generic that commit fb14921 (gpio/mxc: add
> missing initialization of basic_mmio_gpio shadow variables) manged to
> fix in gpio-mxc driver, so that other platform specific drivers do not
> suffer from the same problem over and over again.
>
> Signed-off-by: Shawn Guo <[email protected]>
> ---
> drivers/gpio/gpio-generic.c | 4 ++++
> 1 files changed, 4 insertions(+), 0 deletions(-)
>
> diff --git a/drivers/gpio/gpio-generic.c b/drivers/gpio/gpio-generic.c
> index e38dd0c..cfc9439 100644
> --- a/drivers/gpio/gpio-generic.c
> +++ b/drivers/gpio/gpio-generic.c
> @@ -394,6 +394,10 @@ int bgpio_init(struct bgpio_chip *bgc, struct device *dev,
> return ret;
>
> bgc->data = bgc->read_reg(bgc->reg_dat);
> + if (bgc->gc.set == bgpio_set_set)
> + bgc->data = bgc->read_reg(bgc->reg_set);
> + if (bgc->reg_dir)
> + bgc->dir = bgc->read_reg(bgc->reg_dir);
>
> return ret;
> }
This change is fine for ep93xx.
Acked-by: H Hartley Sweeten <[email protected]>
Hi Grant, Linus,
Can we manage to get it in with the upcoming merge window, as it fixes
a bug that we may not want to live with for another cycle?
Regards,
Shawn
On Thu, May 17, 2012 at 03:35:02PM +0800, Shawn Guo wrote:
> It fixes the issue in gpio-generic that commit fb14921 (gpio/mxc: add
> missing initialization of basic_mmio_gpio shadow variables) manged to
> fix in gpio-mxc driver, so that other platform specific drivers do not
> suffer from the same problem over and over again.
>
> Signed-off-by: Shawn Guo <[email protected]>
> ---
> drivers/gpio/gpio-generic.c | 4 ++++
> 1 files changed, 4 insertions(+), 0 deletions(-)
>
> diff --git a/drivers/gpio/gpio-generic.c b/drivers/gpio/gpio-generic.c
> index e38dd0c..cfc9439 100644
> --- a/drivers/gpio/gpio-generic.c
> +++ b/drivers/gpio/gpio-generic.c
> @@ -394,6 +394,10 @@ int bgpio_init(struct bgpio_chip *bgc, struct device *dev,
> return ret;
>
> bgc->data = bgc->read_reg(bgc->reg_dat);
> + if (bgc->gc.set == bgpio_set_set)
> + bgc->data = bgc->read_reg(bgc->reg_set);
> + if (bgc->reg_dir)
> + bgc->dir = bgc->read_reg(bgc->reg_dir);
>
> return ret;
> }
> --
> 1.7.5.4
>
On Fri, May 18, 2012 at 11:44 PM, Grant Likely
<[email protected]> wrote:
> On Thu, 17 May 2012 15:35:02 +0800, Shawn Guo <[email protected]> wrote:
>> It fixes the issue in gpio-generic that commit fb14921 (gpio/mxc: add
>> missing initialization of basic_mmio_gpio shadow variables) manged to
>> fix in gpio-mxc driver, so that other platform specific drivers do not
>> suffer from the same problem over and over again.
>>
>> Signed-off-by: Shawn Guo <[email protected]>
>> ---
>> ?drivers/gpio/gpio-generic.c | ? ?4 ++++
>> ?1 files changed, 4 insertions(+), 0 deletions(-)
>>
>> diff --git a/drivers/gpio/gpio-generic.c b/drivers/gpio/gpio-generic.c
>> index e38dd0c..cfc9439 100644
>> --- a/drivers/gpio/gpio-generic.c
>> +++ b/drivers/gpio/gpio-generic.c
>> @@ -394,6 +394,10 @@ int bgpio_init(struct bgpio_chip *bgc, struct device *dev,
>> ? ? ? ? ? ? ? return ret;
>>
>> ? ? ? bgc->data = bgc->read_reg(bgc->reg_dat);
>> + ? ? if (bgc->gc.set == bgpio_set_set)
>> + ? ? ? ? ? ? bgc->data = bgc->read_reg(bgc->reg_set);
>> + ? ? if (bgc->reg_dir)
>> + ? ? ? ? ? ? bgc->dir = bgc->read_reg(bgc->reg_dir);
>
> This assumes that the set and dir registers are actually readable
> which isn't the case on some hardware. ?There needs to be a mechanism
> for drivers using bgpio_init to control how data & dir are initialized
> (possibly with some flags; maybe replace the big_endian arg with a
> flags arg).
If you can turn it around quickly, I'll still try to get it in for v3.5
g.