Re: [PATCH] regulator: Allow for asymmetric settling times

[Matthias Kaehlcke]

Hi Mark and Laxman,

thanks for your comments.

El Sun, Apr 30, 2017 at 09:30:01PM +0900 Mark Brown ha dit:

> On Sat, Apr 29, 2017 at 01:32:09PM +0530, Laxman Dewangan wrote:
> > On Saturday 29 April 2017 05:36 AM, Matthias Kaehlcke wrote:
> 
> > > -- regulator-settling-time-us: Settling time, in microseconds, for voltage
> > > -  change if regulator have the constant time for any level voltage 
> > > change.
> > > -  This is useful when regulator have exponential voltage change.
> > > +- regulator-settling-time-up-us: Settling time, in microseconds, for 
> > > voltage
> > > +  increase if the regulator needs a constant time to settle after voltage
> > > +  increases of any level. This is useful for regulators with exponential
> > > +  voltage changes.
> > > +- regulator-settling-time-down-us: Settling time, in microseconds, for 
> > > voltage
> > > +  decrease if the regulator needs a constant time to settle after voltage
> > > +  decreases of any level. This is useful for regulators with exponential
> > > +  voltage changes.
> 
> > Can we have regulator-settling-time-us also so if it is there then up/down
> > same.
> > If up/down different then separate properties can be used.
> 
> Removing the existing binding would also break existing DTs using it
> which we obviously don't want.

I thought it might be ok in this case since the driver just landed.

> I don't see any reason to even deprecate it, like Laxman says it's
> nice and convenient for people with symmetric performance.

Ok, I was 'concerned' about the redundancy, but if that is not seen as
an issue I'm happy to just add the new properties. I'll send an
updated patch soon.

Thanks

Matthias

Re: [PATCH] regulator: Allow for asymmetric settling times

[Matthias Kaehlcke]

Hi Mark and Laxman,

thanks for your comments.

El Sun, Apr 30, 2017 at 09:30:01PM +0900 Mark Brown ha dit:

> On Sat, Apr 29, 2017 at 01:32:09PM +0530, Laxman Dewangan wrote:
> > On Saturday 29 April 2017 05:36 AM, Matthias Kaehlcke wrote:
> 
> > > -- regulator-settling-time-us: Settling time, in microseconds, for voltage
> > > -  change if regulator have the constant time for any level voltage 
> > > change.
> > > -  This is useful when regulator have exponential voltage change.
> > > +- regulator-settling-time-up-us: Settling time, in microseconds, for 
> > > voltage
> > > +  increase if the regulator needs a constant time to settle after voltage
> > > +  increases of any level. This is useful for regulators with exponential
> > > +  voltage changes.
> > > +- regulator-settling-time-down-us: Settling time, in microseconds, for 
> > > voltage
> > > +  decrease if the regulator needs a constant time to settle after voltage
> > > +  decreases of any level. This is useful for regulators with exponential
> > > +  voltage changes.
> 
> > Can we have regulator-settling-time-us also so if it is there then up/down
> > same.
> > If up/down different then separate properties can be used.
> 
> Removing the existing binding would also break existing DTs using it
> which we obviously don't want.

I thought it might be ok in this case since the driver just landed.

> I don't see any reason to even deprecate it, like Laxman says it's
> nice and convenient for people with symmetric performance.

Ok, I was 'concerned' about the redundancy, but if that is not seen as
an issue I'm happy to just add the new properties. I'll send an
updated patch soon.

Thanks

Matthias

Re: [PATCH] regulator: Allow for asymmetric settling times

[Mark Brown]

On Sat, Apr 29, 2017 at 01:32:09PM +0530, Laxman Dewangan wrote:
> On Saturday 29 April 2017 05:36 AM, Matthias Kaehlcke wrote:

> > -- regulator-settling-time-us: Settling time, in microseconds, for voltage
> > -  change if regulator have the constant time for any level voltage change.
> > -  This is useful when regulator have exponential voltage change.
> > +- regulator-settling-time-up-us: Settling time, in microseconds, for 
> > voltage
> > +  increase if the regulator needs a constant time to settle after voltage
> > +  increases of any level. This is useful for regulators with exponential
> > +  voltage changes.
> > +- regulator-settling-time-down-us: Settling time, in microseconds, for 
> > voltage
> > +  decrease if the regulator needs a constant time to settle after voltage
> > +  decreases of any level. This is useful for regulators with exponential
> > +  voltage changes.

> Can we have regulator-settling-time-us also so if it is there then up/down
> same.
> If up/down different then separate properties can be used.

Removing the existing binding would also break existing DTs using it
which we obviously don't want.  I don't see any reason to even deprecate
it, like Laxman says it's nice and convenient for people with symmetric
performance.


signature.asc
Description: PGP signature

Re: [PATCH] regulator: Allow for asymmetric settling times

[Mark Brown]

On Sat, Apr 29, 2017 at 01:32:09PM +0530, Laxman Dewangan wrote:
> On Saturday 29 April 2017 05:36 AM, Matthias Kaehlcke wrote:

> > -- regulator-settling-time-us: Settling time, in microseconds, for voltage
> > -  change if regulator have the constant time for any level voltage change.
> > -  This is useful when regulator have exponential voltage change.
> > +- regulator-settling-time-up-us: Settling time, in microseconds, for 
> > voltage
> > +  increase if the regulator needs a constant time to settle after voltage
> > +  increases of any level. This is useful for regulators with exponential
> > +  voltage changes.
> > +- regulator-settling-time-down-us: Settling time, in microseconds, for 
> > voltage
> > +  decrease if the regulator needs a constant time to settle after voltage
> > +  decreases of any level. This is useful for regulators with exponential
> > +  voltage changes.

> Can we have regulator-settling-time-us also so if it is there then up/down
> same.
> If up/down different then separate properties can be used.

Removing the existing binding would also break existing DTs using it
which we obviously don't want.  I don't see any reason to even deprecate
it, like Laxman says it's nice and convenient for people with symmetric
performance.


signature.asc
Description: PGP signature

Re: [PATCH] regulator: Allow for asymmetric settling times

[Laxman Dewangan]

On Saturday 29 April 2017 05:36 AM, Matthias Kaehlcke wrote:

Some regulators have different settling times for voltage increases and
decreases. To avoid a time penalty on the faster transition extend the
settling time property to allow for different settings for upward and
downward transitions.

Signed-off-by: Matthias Kaehlcke 
---
Dependencies (from broonie/regulator topic/settle):
  - regulator: DT: Add settling time property for non-linear voltage change
  - regulator: Add settling time for non-linear voltage transition

Sorry for not bringing this up during the review of the 'settling time'
patch, I just came across it when looking to revive a similar change I
sent out some time ago (https://patchwork.kernel.org/patch/9332051/).

  Documentation/devicetree/bindings/regulator/regulator.txt | 11 ---
  drivers/regulator/core.c  |  8 ++--
  drivers/regulator/of_regulator.c  |  9 +++--
  include/linux/regulator/machine.h |  9 ++---
  4 files changed, 27 insertions(+), 10 deletions(-)

I think DT change and code change go in different patches.


diff --git a/Documentation/devicetree/bindings/regulator/regulator.txt 
b/Documentation/devicetree/bindings/regulator/regulator.txt
index d18edb075e1c..f21fead1c802 100644
--- a/Documentation/devicetree/bindings/regulator/regulator.txt
+++ b/Documentation/devicetree/bindings/regulator/regulator.txt
@@ -21,9 +21,14 @@ Optional properties:
design requires. This property describes the total system ramp time
required due to the combination of internal ramping of the regulator itself,
and board design issues such as trace capacitance and load on the supply.
-- regulator-settling-time-us: Settling time, in microseconds, for voltage
-  change if regulator have the constant time for any level voltage change.
-  This is useful when regulator have exponential voltage change.
+- regulator-settling-time-up-us: Settling time, in microseconds, for voltage
+  increase if the regulator needs a constant time to settle after voltage
+  increases of any level. This is useful for regulators with exponential
+  voltage changes.
+- regulator-settling-time-down-us: Settling time, in microseconds, for voltage
+  decrease if the regulator needs a constant time to settle after voltage
+  decreases of any level. This is useful for regulators with exponential
+  voltage changes.


Can we have regulator-settling-time-us also so if it is there then 
up/down same.

If up/down different then separate properties can be used.


Also in driver, if up/dn are not provided and only 
regulator-settling-time-us is provided then up/dn can take value from 
this property.

Re: [PATCH] regulator: Allow for asymmetric settling times

[Laxman Dewangan]

On Saturday 29 April 2017 05:36 AM, Matthias Kaehlcke wrote:

Some regulators have different settling times for voltage increases and
decreases. To avoid a time penalty on the faster transition extend the
settling time property to allow for different settings for upward and
downward transitions.

Signed-off-by: Matthias Kaehlcke 
---
Dependencies (from broonie/regulator topic/settle):
  - regulator: DT: Add settling time property for non-linear voltage change
  - regulator: Add settling time for non-linear voltage transition

Sorry for not bringing this up during the review of the 'settling time'
patch, I just came across it when looking to revive a similar change I
sent out some time ago (https://patchwork.kernel.org/patch/9332051/).

  Documentation/devicetree/bindings/regulator/regulator.txt | 11 ---
  drivers/regulator/core.c  |  8 ++--
  drivers/regulator/of_regulator.c  |  9 +++--
  include/linux/regulator/machine.h |  9 ++---
  4 files changed, 27 insertions(+), 10 deletions(-)

I think DT change and code change go in different patches.


diff --git a/Documentation/devicetree/bindings/regulator/regulator.txt 
b/Documentation/devicetree/bindings/regulator/regulator.txt
index d18edb075e1c..f21fead1c802 100644
--- a/Documentation/devicetree/bindings/regulator/regulator.txt
+++ b/Documentation/devicetree/bindings/regulator/regulator.txt
@@ -21,9 +21,14 @@ Optional properties:
design requires. This property describes the total system ramp time
required due to the combination of internal ramping of the regulator itself,
and board design issues such as trace capacitance and load on the supply.
-- regulator-settling-time-us: Settling time, in microseconds, for voltage
-  change if regulator have the constant time for any level voltage change.
-  This is useful when regulator have exponential voltage change.
+- regulator-settling-time-up-us: Settling time, in microseconds, for voltage
+  increase if the regulator needs a constant time to settle after voltage
+  increases of any level. This is useful for regulators with exponential
+  voltage changes.
+- regulator-settling-time-down-us: Settling time, in microseconds, for voltage
+  decrease if the regulator needs a constant time to settle after voltage
+  decreases of any level. This is useful for regulators with exponential
+  voltage changes.


Can we have regulator-settling-time-us also so if it is there then 
up/down same.

If up/down different then separate properties can be used.


Also in driver, if up/dn are not provided and only 
regulator-settling-time-us is provided then up/dn can take value from 
this property.

[PATCH] regulator: Allow for asymmetric settling times

[Matthias Kaehlcke]

Some regulators have different settling times for voltage increases and
decreases. To avoid a time penalty on the faster transition extend the
settling time property to allow for different settings for upward and
downward transitions.

Signed-off-by: Matthias Kaehlcke 
---
Dependencies (from broonie/regulator topic/settle):
 - regulator: DT: Add settling time property for non-linear voltage change
 - regulator: Add settling time for non-linear voltage transition

Sorry for not bringing this up during the review of the 'settling time'
patch, I just came across it when looking to revive a similar change I
sent out some time ago (https://patchwork.kernel.org/patch/9332051/).

 Documentation/devicetree/bindings/regulator/regulator.txt | 11 ---
 drivers/regulator/core.c  |  8 ++--
 drivers/regulator/of_regulator.c  |  9 +++--
 include/linux/regulator/machine.h |  9 ++---
 4 files changed, 27 insertions(+), 10 deletions(-)

diff --git a/Documentation/devicetree/bindings/regulator/regulator.txt 
b/Documentation/devicetree/bindings/regulator/regulator.txt
index d18edb075e1c..f21fead1c802 100644
--- a/Documentation/devicetree/bindings/regulator/regulator.txt
+++ b/Documentation/devicetree/bindings/regulator/regulator.txt
@@ -21,9 +21,14 @@ Optional properties:
   design requires. This property describes the total system ramp time
   required due to the combination of internal ramping of the regulator itself,
   and board design issues such as trace capacitance and load on the supply.
-- regulator-settling-time-us: Settling time, in microseconds, for voltage
-  change if regulator have the constant time for any level voltage change.
-  This is useful when regulator have exponential voltage change.
+- regulator-settling-time-up-us: Settling time, in microseconds, for voltage
+  increase if the regulator needs a constant time to settle after voltage
+  increases of any level. This is useful for regulators with exponential
+  voltage changes.
+- regulator-settling-time-down-us: Settling time, in microseconds, for voltage
+  decrease if the regulator needs a constant time to settle after voltage
+  decreases of any level. This is useful for regulators with exponential
+  voltage changes.
 - regulator-soft-start: Enable soft start so that voltage ramps slowly
 - regulator-state-mem sub-root node for Suspend-to-RAM mode
   : suspend to memory, the device goes to sleep, but all data stored in memory,
diff --git a/drivers/regulator/core.c b/drivers/regulator/core.c
index 811096b23143..4df86c0da123 100644
--- a/drivers/regulator/core.c
+++ b/drivers/regulator/core.c
@@ -2773,8 +2773,12 @@ static int _regulator_set_voltage_time(struct 
regulator_dev *rdev,
ramp_delay = rdev->constraints->ramp_delay;
else if (rdev->desc->ramp_delay)
ramp_delay = rdev->desc->ramp_delay;
-   else if (rdev->constraints->settling_time)
-   return rdev->constraints->settling_time;
+   else if (rdev->constraints->settling_time_up &&
+(new_uV > old_uV))
+   return rdev->constraints->settling_time_up;
+   else if (rdev->constraints->settling_time_down &&
+(new_uV < old_uV))
+   return rdev->constraints->settling_time_down;
 
if (ramp_delay == 0) {
rdev_dbg(rdev, "ramp_delay not set\n");
diff --git a/drivers/regulator/of_regulator.c b/drivers/regulator/of_regulator.c
index 09d677d5d3f0..4d36c0e4c9b4 100644
--- a/drivers/regulator/of_regulator.c
+++ b/drivers/regulator/of_regulator.c
@@ -86,9 +86,14 @@ static void of_get_regulation_constraints(struct device_node 
*np,
constraints->ramp_disable = true;
}
 
-   ret = of_property_read_u32(np, "regulator-settling-time-us", );
+   ret = of_property_read_u32(np, "regulator-settling-time-up-us", );
if (!ret)
-   constraints->settling_time = pval;
+   constraints->settling_time_up = pval;
+
+   ret = of_property_read_u32(np, "regulator-settling-time-down-us",
+  );
+   if (!ret)
+   constraints->settling_time_down = pval;
 
ret = of_property_read_u32(np, "regulator-enable-ramp-delay", );
if (!ret)
diff --git a/include/linux/regulator/machine.h 
b/include/linux/regulator/machine.h
index 117699d1f7df..a92829e86b5d 100644
--- a/include/linux/regulator/machine.h
+++ b/include/linux/regulator/machine.h
@@ -108,8 +108,10 @@ struct regulator_state {
  * @initial_state: Suspend state to set by default.
  * @initial_mode: Mode to set at startup.
  * @ramp_delay: Time to settle down after voltage change (unit: uV/us)
- * @settling_time: Time to settle down after voltage change when voltage
- *change is non-linear (unit: microseconds).
+ * @settling_time_up: Time to settle down after voltage increase when 

[PATCH] regulator: Allow for asymmetric settling times

[Matthias Kaehlcke]

Some regulators have different settling times for voltage increases and
decreases. To avoid a time penalty on the faster transition extend the
settling time property to allow for different settings for upward and
downward transitions.

Signed-off-by: Matthias Kaehlcke 
---
Dependencies (from broonie/regulator topic/settle):
 - regulator: DT: Add settling time property for non-linear voltage change
 - regulator: Add settling time for non-linear voltage transition

Sorry for not bringing this up during the review of the 'settling time'
patch, I just came across it when looking to revive a similar change I
sent out some time ago (https://patchwork.kernel.org/patch/9332051/).

 Documentation/devicetree/bindings/regulator/regulator.txt | 11 ---
 drivers/regulator/core.c  |  8 ++--
 drivers/regulator/of_regulator.c  |  9 +++--
 include/linux/regulator/machine.h |  9 ++---
 4 files changed, 27 insertions(+), 10 deletions(-)

diff --git a/Documentation/devicetree/bindings/regulator/regulator.txt 
b/Documentation/devicetree/bindings/regulator/regulator.txt
index d18edb075e1c..f21fead1c802 100644
--- a/Documentation/devicetree/bindings/regulator/regulator.txt
+++ b/Documentation/devicetree/bindings/regulator/regulator.txt
@@ -21,9 +21,14 @@ Optional properties:
   design requires. This property describes the total system ramp time
   required due to the combination of internal ramping of the regulator itself,
   and board design issues such as trace capacitance and load on the supply.
-- regulator-settling-time-us: Settling time, in microseconds, for voltage
-  change if regulator have the constant time for any level voltage change.
-  This is useful when regulator have exponential voltage change.
+- regulator-settling-time-up-us: Settling time, in microseconds, for voltage
+  increase if the regulator needs a constant time to settle after voltage
+  increases of any level. This is useful for regulators with exponential
+  voltage changes.
+- regulator-settling-time-down-us: Settling time, in microseconds, for voltage
+  decrease if the regulator needs a constant time to settle after voltage
+  decreases of any level. This is useful for regulators with exponential
+  voltage changes.
 - regulator-soft-start: Enable soft start so that voltage ramps slowly
 - regulator-state-mem sub-root node for Suspend-to-RAM mode
   : suspend to memory, the device goes to sleep, but all data stored in memory,
diff --git a/drivers/regulator/core.c b/drivers/regulator/core.c
index 811096b23143..4df86c0da123 100644
--- a/drivers/regulator/core.c
+++ b/drivers/regulator/core.c
@@ -2773,8 +2773,12 @@ static int _regulator_set_voltage_time(struct 
regulator_dev *rdev,
ramp_delay = rdev->constraints->ramp_delay;
else if (rdev->desc->ramp_delay)
ramp_delay = rdev->desc->ramp_delay;
-   else if (rdev->constraints->settling_time)
-   return rdev->constraints->settling_time;
+   else if (rdev->constraints->settling_time_up &&
+(new_uV > old_uV))
+   return rdev->constraints->settling_time_up;
+   else if (rdev->constraints->settling_time_down &&
+(new_uV < old_uV))
+   return rdev->constraints->settling_time_down;
 
if (ramp_delay == 0) {
rdev_dbg(rdev, "ramp_delay not set\n");
diff --git a/drivers/regulator/of_regulator.c b/drivers/regulator/of_regulator.c
index 09d677d5d3f0..4d36c0e4c9b4 100644
--- a/drivers/regulator/of_regulator.c
+++ b/drivers/regulator/of_regulator.c
@@ -86,9 +86,14 @@ static void of_get_regulation_constraints(struct device_node 
*np,
constraints->ramp_disable = true;
}
 
-   ret = of_property_read_u32(np, "regulator-settling-time-us", );
+   ret = of_property_read_u32(np, "regulator-settling-time-up-us", );
if (!ret)
-   constraints->settling_time = pval;
+   constraints->settling_time_up = pval;
+
+   ret = of_property_read_u32(np, "regulator-settling-time-down-us",
+  );
+   if (!ret)
+   constraints->settling_time_down = pval;
 
ret = of_property_read_u32(np, "regulator-enable-ramp-delay", );
if (!ret)
diff --git a/include/linux/regulator/machine.h 
b/include/linux/regulator/machine.h
index 117699d1f7df..a92829e86b5d 100644
--- a/include/linux/regulator/machine.h
+++ b/include/linux/regulator/machine.h
@@ -108,8 +108,10 @@ struct regulator_state {
  * @initial_state: Suspend state to set by default.
  * @initial_mode: Mode to set at startup.
  * @ramp_delay: Time to settle down after voltage change (unit: uV/us)
- * @settling_time: Time to settle down after voltage change when voltage
- *change is non-linear (unit: microseconds).
+ * @settling_time_up: Time to settle down after voltage increase when voltage
+ *