2010-08-28 17:16:03

by Gregory Bean

[permalink] [raw]
Subject: [PATCH] gpio: sx150x: correct and refine reset-on-probe behavior.

Replace the arbitrary software-reset call from the device-probe
method, because:
- It is defective. To work correctly, it should be two byte writes,
not a single word write. As it stands, it does nothing.
- Some devices with sx150x expanders installed have their NRESET pins
ganged on the same line, so resetting one causes the others to reset -
not a nice thing to do arbitrarily!
- The probe, usually taking place at boot, implies a recent hard-reset,
so a software reset at this point is just a waste of energy anyway.

Therefore, make it optional, defaulting to off, as this will match the
common case of probing at powerup and also matches the current
broken no-op behavior.

Signed-off-by: Gregory Bean <[email protected]>
---
drivers/gpio/sx150x.c | 29 ++++++++++++++++++++++++-----
include/linux/i2c/sx150x.h | 4 ++++
2 files changed, 28 insertions(+), 5 deletions(-)

diff --git a/drivers/gpio/sx150x.c b/drivers/gpio/sx150x.c
index b42f42c..6d68018 100644
--- a/drivers/gpio/sx150x.c
+++ b/drivers/gpio/sx150x.c
@@ -459,17 +459,36 @@ static int sx150x_init_io(struct sx150x_chip *chip, u8 base, u16 cfg)
return err;
}

-static int sx150x_init_hw(struct sx150x_chip *chip,
- struct sx150x_platform_data *pdata)
+static int sx150x_reset(struct sx150x_chip *chip)
{
- int err = 0;
+ int err;
+
+ err = i2c_smbus_write_byte_data(chip->client,
+ chip->dev_cfg->reg_reset,
+ 0x12);
+ if (err < 0)
+ return err;

- err = i2c_smbus_write_word_data(chip->client,
+ err = i2c_smbus_write_byte_data(chip->client,
chip->dev_cfg->reg_reset,
- 0x3412);
+ 0x34);
if (err < 0)
return err;

+ return 0;
+}
+
+static int sx150x_init_hw(struct sx150x_chip *chip,
+ struct sx150x_platform_data *pdata)
+{
+ int err = 0;
+
+ if (pdata->reset_during_probe) {
+ err = sx150x_reset(chip);
+ if (err < 0)
+ return err;
+ }
+
err = sx150x_i2c_write(chip->client,
chip->dev_cfg->reg_misc,
0x01);
diff --git a/include/linux/i2c/sx150x.h b/include/linux/i2c/sx150x.h
index ee3049c..52baa79 100644
--- a/include/linux/i2c/sx150x.h
+++ b/include/linux/i2c/sx150x.h
@@ -63,6 +63,9 @@
* IRQ lines will appear. Similarly to gpio_base, the expander
* will create a block of irqs beginning at this number.
* This value is ignored if irq_summary is < 0.
+ * @reset_during_probe: If set to true, the driver will trigger a full
+ * reset of the chip at the beginning of the probe
+ * in order to place it in a known state.
*/
struct sx150x_platform_data {
unsigned gpio_base;
@@ -73,6 +76,7 @@ struct sx150x_platform_data {
u16 io_polarity;
int irq_summary;
unsigned irq_base;
+ bool reset_during_probe;
};

#endif /* __LINUX_I2C_SX150X_H */
--
1.7.0.4

--
Employee of Qualcomm Innovation Center, Inc.
Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.


2010-08-29 09:43:56

by Jean Delvare

[permalink] [raw]
Subject: Re: [PATCH] gpio: sx150x: correct and refine reset-on-probe behavior.

Hi Gregory,

On Sat, 28 Aug 2010 10:15:52 -0700, Gregory Bean wrote:
> Replace the arbitrary software-reset call from the device-probe
> method, because:
> - It is defective. To work correctly, it should be two byte writes,
> not a single word write. As it stands, it does nothing.
> - Some devices with sx150x expanders installed have their NRESET pins
> ganged on the same line, so resetting one causes the others to reset -
> not a nice thing to do arbitrarily!
> - The probe, usually taking place at boot, implies a recent hard-reset,
> so a software reset at this point is just a waste of energy anyway.
>
> Therefore, make it optional, defaulting to off, as this will match the
> common case of probing at powerup and also matches the current
> broken no-op behavior.
>
> Signed-off-by: Gregory Bean <[email protected]>
> ---
> drivers/gpio/sx150x.c | 29 ++++++++++++++++++++++++-----
> include/linux/i2c/sx150x.h | 4 ++++
> 2 files changed, 28 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/gpio/sx150x.c b/drivers/gpio/sx150x.c
> index b42f42c..6d68018 100644
> --- a/drivers/gpio/sx150x.c
> +++ b/drivers/gpio/sx150x.c
> @@ -459,17 +459,36 @@ static int sx150x_init_io(struct sx150x_chip *chip, u8 base, u16 cfg)
> return err;
> }
>
> -static int sx150x_init_hw(struct sx150x_chip *chip,
> - struct sx150x_platform_data *pdata)
> +static int sx150x_reset(struct sx150x_chip *chip)
> {
> - int err = 0;
> + int err;
> +
> + err = i2c_smbus_write_byte_data(chip->client,
> + chip->dev_cfg->reg_reset,
> + 0x12);
> + if (err < 0)
> + return err;
>
> - err = i2c_smbus_write_word_data(chip->client,
> + err = i2c_smbus_write_byte_data(chip->client,
> chip->dev_cfg->reg_reset,
> - 0x3412);
> + 0x34);
> if (err < 0)
> return err;
>
> + return 0;

You might as well just "return err" in both cases.

> +}
> +
> +static int sx150x_init_hw(struct sx150x_chip *chip,
> + struct sx150x_platform_data *pdata)
> +{
> + int err = 0;
> +
> + if (pdata->reset_during_probe) {
> + err = sx150x_reset(chip);
> + if (err < 0)
> + return err;
> + }
> +
> err = sx150x_i2c_write(chip->client,
> chip->dev_cfg->reg_misc,
> 0x01);
> diff --git a/include/linux/i2c/sx150x.h b/include/linux/i2c/sx150x.h
> index ee3049c..52baa79 100644
> --- a/include/linux/i2c/sx150x.h
> +++ b/include/linux/i2c/sx150x.h
> @@ -63,6 +63,9 @@
> * IRQ lines will appear. Similarly to gpio_base, the expander
> * will create a block of irqs beginning at this number.
> * This value is ignored if irq_summary is < 0.
> + * @reset_during_probe: If set to true, the driver will trigger a full
> + * reset of the chip at the beginning of the probe
> + * in order to place it in a known state.
> */
> struct sx150x_platform_data {
> unsigned gpio_base;
> @@ -73,6 +76,7 @@ struct sx150x_platform_data {
> u16 io_polarity;
> int irq_summary;
> unsigned irq_base;
> + bool reset_during_probe;
> };
>
> #endif /* __LINUX_I2C_SX150X_H */


--
Jean Delvare