The DA9062 can be the device used to power the CPU. In that case, it can
be used to power off the system. In the CONTROL_A register, the M_*_EN
bits must be zero for the corresponding *_EN bits to have an effect. We
zero them all to turn off the system.
Signed-off-by: Helmut Grohne <[email protected]>
---
drivers/mfd/da9062-core.c | 31 +++++++++++++++++++++++++++++++
1 file changed, 31 insertions(+)
This series effectively is a rebased resend. The earlier posting was
https://lore.kernel.org/all/20200107120559.GA700@laureti-dev/. At that time,
Adam Thomson critisized the use of regmap and i2c inside a pm_power_off hook
since irqs are disabled. He reached out to Lee Jones, who asked Mark Brown and
Wolfram Sang, but never got any reply. I noted that a fair number of other
drivers do use regmap and i2c despite this issue. In the mean time, more
instances were added:
* drivers/mfd/acer-ec-a500.c uses i2c
* drivers/mfd/ntxec.c uses i2c
* drivers/mfd/rk808.c uses regmap and i2c
Given that we proceeded with accepting the use of i2c and regmap inside
pm_power_off hooks, I think we can proceed with this patch as well.
Helmut
diff --git a/drivers/mfd/da9062-core.c b/drivers/mfd/da9062-core.c
index 2774b2cbaea6..e7af5b5f16e0 100644
--- a/drivers/mfd/da9062-core.c
+++ b/drivers/mfd/da9062-core.c
@@ -620,6 +620,23 @@ static const struct of_device_id da9062_dt_ids[] = {
};
MODULE_DEVICE_TABLE(of, da9062_dt_ids);
+/* Hold client since pm_power_off is global. */
+static struct i2c_client *da9062_i2c_client;
+
+static void da9062_power_off(void)
+{
+ struct da9062 *chip = i2c_get_clientdata(da9062_i2c_client);
+ const unsigned int mask = DA9062AA_SYSTEM_EN_MASK |
+ DA9062AA_POWER_EN_MASK | DA9062AA_POWER1_EN_MASK |
+ DA9062AA_M_SYSTEM_EN_MASK | DA9062AA_M_POWER_EN_MASK |
+ DA9062AA_M_POWER1_EN_MASK;
+ int ret = regmap_update_bits(chip->regmap, DA9062AA_CONTROL_A, mask, 0);
+
+ if (ret < 0)
+ dev_err(&da9062_i2c_client->dev,
+ "DA9062AA_CONTROL_A update failed, %d\n", ret);
+}
+
static int da9062_i2c_probe(struct i2c_client *i2c,
const struct i2c_device_id *id)
{
@@ -720,6 +737,15 @@ static int da9062_i2c_probe(struct i2c_client *i2c,
return ret;
}
+ if (of_device_is_system_power_controller(i2c->dev.of_node)) {
+ if (!pm_power_off) {
+ da9062_i2c_client = i2c;
+ pm_power_off = da9062_power_off;
+ } else {
+ dev_warn(&i2c->dev, "Poweroff callback already assigned\n");
+ }
+ }
+
return ret;
}
@@ -727,6 +753,11 @@ static int da9062_i2c_remove(struct i2c_client *i2c)
{
struct da9062 *chip = i2c_get_clientdata(i2c);
+ if (pm_power_off == da9062_power_off)
+ pm_power_off = NULL;
+ if (da9062_i2c_client)
+ da9062_i2c_client = NULL;
+
mfd_remove_devices(chip->dev);
regmap_del_irq_chip(i2c->irq, chip->regmap_irq);
--
2.20.1
Dipl.-Inf. Helmut Grohne
Research and Development
Application Engineering
Phone: +49 (371) 24354 0 ∙ Fax: +49 (371) 24354 020
[email protected] ∙ https://www.intenta.de
Intenta GmbH ∙ Ahornstraße 55 ∙ 09112 Chemnitz, Germany
Managing Director: Dr.-Ing. Basel Fardi ∙ VAT/USt-IdNr.: DE 275745394
Commercial register: HRB 26404 Amtsgericht Chemnitz
On 22 April 2022 07:17, Helmut Grohne wrote:
> The DA9062 can be the device used to power the CPU. In that case, it can
> be used to power off the system. In the CONTROL_A register, the M_*_EN
> bits must be zero for the corresponding *_EN bits to have an effect. We
> zero them all to turn off the system.
>
> Signed-off-by: Helmut Grohne <[email protected]>
> ---
> drivers/mfd/da9062-core.c | 31 +++++++++++++++++++++++++++++++
> 1 file changed, 31 insertions(+)
>
> This series effectively is a rebased resend. The earlier posting was
> https://lore.kernel.org/all/20200107120559.GA700@laureti-dev/. At that time,
> Adam Thomson critisized the use of regmap and i2c inside a pm_power_off hook
> since irqs are disabled. He reached out to Lee Jones, who asked Mark Brown and
> Wolfram Sang, but never got any reply. I noted that a fair number of other
> drivers do use regmap and i2c despite this issue. In the mean time, more
> instances were added:
> * drivers/mfd/acer-ec-a500.c uses i2c
> * drivers/mfd/ntxec.c uses i2c
> * drivers/mfd/rk808.c uses regmap and i2c
> Given that we proceeded with accepting the use of i2c and regmap inside
> pm_power_off hooks, I think we can proceed with this patch as well.
>
> Helmut
I still have concerns, but maybe it's up to the integrator of a platform to
determine if this approach for poweroff is valid and usable, although I still
think it would be good to have checks somewhere as to whether a platform can
actually perform atmoic I2C access when IRQs are disabled. However, without
further insight right now:
Reviewed-by: Adam Thomson <[email protected]>
On Fri, 22 Apr 2022, Helmut Grohne wrote:
> The DA9062 can be the device used to power the CPU. In that case, it can
> be used to power off the system. In the CONTROL_A register, the M_*_EN
> bits must be zero for the corresponding *_EN bits to have an effect. We
> zero them all to turn off the system.
>
> Signed-off-by: Helmut Grohne <[email protected]>
> ---
> drivers/mfd/da9062-core.c | 31 +++++++++++++++++++++++++++++++
> 1 file changed, 31 insertions(+)
>
> This series effectively is a rebased resend. The earlier posting was
> https://lore.kernel.org/all/20200107120559.GA700@laureti-dev/. At that time,
> Adam Thomson critisized the use of regmap and i2c inside a pm_power_off hook
> since irqs are disabled. He reached out to Lee Jones, who asked Mark Brown and
> Wolfram Sang, but never got any reply. I noted that a fair number of other
> drivers do use regmap and i2c despite this issue. In the mean time, more
> instances were added:
> * drivers/mfd/acer-ec-a500.c uses i2c
> * drivers/mfd/ntxec.c uses i2c
> * drivers/mfd/rk808.c uses regmap and i2c
> Given that we proceeded with accepting the use of i2c and regmap inside
> pm_power_off hooks, I think we can proceed with this patch as well.
>
> Helmut
>
> diff --git a/drivers/mfd/da9062-core.c b/drivers/mfd/da9062-core.c
> index 2774b2cbaea6..e7af5b5f16e0 100644
> --- a/drivers/mfd/da9062-core.c
> +++ b/drivers/mfd/da9062-core.c
> @@ -620,6 +620,23 @@ static const struct of_device_id da9062_dt_ids[] = {
> };
> MODULE_DEVICE_TABLE(of, da9062_dt_ids);
>
> +/* Hold client since pm_power_off is global. */
Please drop this comment.
> +static struct i2c_client *da9062_i2c_client;
> +
> +static void da9062_power_off(void)
> +{
> + struct da9062 *chip = i2c_get_clientdata(da9062_i2c_client);
> + const unsigned int mask = DA9062AA_SYSTEM_EN_MASK |
> + DA9062AA_POWER_EN_MASK | DA9062AA_POWER1_EN_MASK |
> + DA9062AA_M_SYSTEM_EN_MASK | DA9062AA_M_POWER_EN_MASK |
> + DA9062AA_M_POWER1_EN_MASK;
> + int ret = regmap_update_bits(chip->regmap, DA9062AA_CONTROL_A, mask, 0);
This is messy. Please separate declarations and assignments here.
The top one is passable.
> + if (ret < 0)
> + dev_err(&da9062_i2c_client->dev,
> + "DA9062AA_CONTROL_A update failed, %d\n", ret);
You're talking to the user here.
Please use language that is more user-friendly.
> +}
> +
> static int da9062_i2c_probe(struct i2c_client *i2c,
> const struct i2c_device_id *id)
> {
> @@ -720,6 +737,15 @@ static int da9062_i2c_probe(struct i2c_client *i2c,
> return ret;
> }
>
> + if (of_device_is_system_power_controller(i2c->dev.of_node)) {
> + if (!pm_power_off) {
> + da9062_i2c_client = i2c;
> + pm_power_off = da9062_power_off;
> + } else {
> + dev_warn(&i2c->dev, "Poweroff callback already assigned\n");
Do we really mind/care?
Is there anything we can do about it?
Thus, do we really need to warn() about it?
> + }
> + }
> +
> return ret;
> }
>
> @@ -727,6 +753,11 @@ static int da9062_i2c_remove(struct i2c_client *i2c)
> {
> struct da9062 *chip = i2c_get_clientdata(i2c);
>
> + if (pm_power_off == da9062_power_off)
> + pm_power_off = NULL;
> + if (da9062_i2c_client)
> + da9062_i2c_client = NULL;
> +
> mfd_remove_devices(chip->dev);
> regmap_del_irq_chip(i2c->irq, chip->regmap_irq);
>
--
Lee Jones [李琼斯]
Principal Technical Lead - Developer Services
Linaro.org │ Open source software for Arm SoCs
Follow Linaro: Facebook | Twitter | Blog