Skip to contentSkip to conversation

regulator pre-enable

PATCH In review Jianqun Xu 2 participants 8 messages Aug 17, 2021 – Aug 18, 2021

Subsystems (2)
lore
Share & tools
lore

Jianqun Xu

author
#0 lore raw

v2 cover regulator pre-enable

Rockchip io-domain care about regulator pre-enable

Jianqun Xu (3):
regulator: add PRE_ENABLE event define
regulator: core: add pre-enable event notify to regulator
soc: rockchip: io-domain: do more thing about regulator notify

 drivers/regulator/core.c           | 3 +++
 drivers/soc/rockchip/io-domain.c   | 6 ++++++
 include/linux/regulator/consumer.h | 2 ++
 3 files changed, 11 insertions(+)
Signature · 7 lines
v2:
 - remove patch "regulator: core: notify regulator enable with the voltage value"
 - use "uV = regulator_get_voltage(supply->reg);" but from notify data
v1:
 - first

2.25.1
author
#1 lore raw

PATCH v2 1/3 regulator: add PRE_ENABLE event define

Add REGULATOR_EVENT_PRE_ENABLE to allow to notify driver that the regulator is about to enabled.

---
v2: none

 include/linux/regulator/consumer.h | 2 ++
 1 file changed, 2 insertions(+)
include/linux/regulator/consumer.h +2 -0
index 20e84a84fb77..edd31f0dad17 100644
@@ -102,6 +102,7 @@ struct regulator_dev;
  *                      Data passed is old voltage cast to (void *).
  * PRE_DISABLE    Regulator is about to be disabled
  * ABORT_DISABLE  Regulator disable failed for some reason
+ * PRE_ENABLE     Regulator is about to be enabled
  *
  * NOTE: These events can be OR'ed together when passed into handler.
  */
@@ -119,6 +120,7 @@ struct regulator_dev;
 #define REGULATOR_EVENT_PRE_DISABLE		0x400
 #define REGULATOR_EVENT_ABORT_DISABLE		0x800
 #define REGULATOR_EVENT_ENABLE			0x1000
+#define REGULATOR_EVENT_PRE_ENABLE		0x2000
 
 /*
  * Regulator errors that can be queried using regulator_get_error_flags
Signature · 1 line
2.25.1
maintainer
#2 lore raw

Re: [PATCH v2 1/3] regulator: add PRE_ENABLE event define

Quoted 2 lines · Jianqun Xu

Add REGULATOR_EVENT_PRE_ENABLE to allow to notify driver that the regulator is about to enabled.

This doesn't apply against current code, please check and resend.

author
#3 lore raw

PATCH v2 2/3 regulator: core: add pre-enable event notify to regulator

In reply to Jianqun Xu

Notify the event about regulator to be enabled to driver.

The IO-DOMAIN driver on Rockchip SoCs will take this event and then to configure the io-domain mode before regualtor_enable.

---
v2: none

 drivers/regulator/core.c | 3 +++
 1 file changed, 3 insertions(+)
drivers/regulator/core.c +3 -0
index f192bf19492e..d7dacc608509 100644
@@ -2654,6 +2654,9 @@ static int _regulator_enable(struct regulator *regulator)
 				goto err_consumer_disable;
 			}
 
+			_notifier_call_chain(rdev, REGULATOR_EVENT_PRE_ENABLE,
+					     NULL);
+
 			ret = _regulator_do_enable(rdev);
 			if (ret < 0)
 				goto err_consumer_disable;
Signature · 1 line
2.25.1
author
#4 lore raw

PATCH v2 3/3 soc: rockchip: io-domain: do more thing about regulator notify

In reply to Jianqun Xu

Do a fix to rockchip io-domain, follow this orders:

* system running state
-> io-domain vsel to 3.3V

    -> regulator_enable
      -> vsel change according to regulator voltage

* system running state
-> io-domain vsel to 3.3V

    -> regulator_disable

Found on some Rockchip SoCs, the regulator enable or disable without care about the io-domain maybe caused soc damaged.

---
v2:
- use "uV = regulator_get_voltage(supply->reg);" but from notify data
v1:
- first version

 drivers/soc/rockchip/io-domain.c | 6 ++++++
 1 file changed, 6 insertions(+)
drivers/soc/rockchip/io-domain.c +6 -0
index cf8182fc3642..3c59077fafb1 100644
@@ -123,6 +123,12 @@ static int rockchip_iodomain_notify(struct notifier_block *nb,
 	} else if (event & (REGULATOR_EVENT_VOLTAGE_CHANGE |
 			    REGULATOR_EVENT_ABORT_VOLTAGE_CHANGE)) {
 		uV = (unsigned long)data;
+	} else if (event & REGULATOR_EVENT_PRE_ENABLE) {
+		uV = MAX_VOLTAGE_3_3;
+	} else if (event & REGULATOR_EVENT_PRE_DISABLE) {
+		uV = MAX_VOLTAGE_3_3;
+	} else if (event & REGULATOR_EVENT_ENABLE) {
+		uV = regulator_get_voltage(supply->reg);
 	} else {
 		return NOTIFY_OK;
 	}
Signature · 1 line
2.25.1
maintainer
#5 lore raw

Re: [PATCH v2 3/3] soc: rockchip: io-domain: do more thing about regulator notify

drivers/soc/rockchip/io-domain.c static int rockchip_iodomain_notify(struct notifier_block *nb, · line 132

Quoted 3 lines · Jianqun Xu
drivers/soc/rockchip/io-domain.c +2 -0
+	} else if (event & REGULATOR_EVENT_ENABLE) {
+		uV = regulator_get_voltage(supply->reg);
 	} else {

I am very surprised this doesn't cause locking issues given that we might call notifiers with the regulator API's locks held. Have you tested this with lockdep on?

#6 lore raw

Re: Re: [PATCH v2 3/3] soc: rockchip: io-domain: do more thing about regulator notify

Hi Mark
--------------
[email protected]

Quoted 9 lines · Jianqun Xu
Quoted 4 lines
drivers/soc/rockchip/io-domain.c +2 -0
+	} else if (event & REGULATOR_EVENT_ENABLE) {
+	uV = regulator_get_voltage(supply->reg);

  } else {

I am very surprised this doesn't cause locking issues given that we might call notifiers with the regulator API's locks held.  Have you tested this with lockdep on?

Thanks for your reply, there really has locking issue, our test team fail to find out it but get a pass result.

So, if the voltage cannot get here by driver, can the notify  pass it, like pre-version patch?
_______________________________________________
Linux-rockchip mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-rockchip

#7 lore raw

Re: Re: [PATCH v2 3/3] soc: rockchip: io-domain: do more thing about regulator notify

Hi Mark
--------------
[email protected]

Quoted 12 lines

Hi Mark
--------------
[email protected]

Quoted 9 lines · Jianqun Xu
Quoted 4 lines
drivers/soc/rockchip/io-domain.c +2 -0
+	} else if (event & REGULATOR_EVENT_ENABLE) {
+	uV = regulator_get_voltage(supply->reg);

  } else {

I am very surprised this doesn't cause locking issues given that we might call notifiers with the regulator API's locks held.  Have you tested this with lockdep on?

Quoted 4 lines

Thanks for your reply, there really has locking issue, our test team fail to find out it but get a pass result.

So, if the voltage cannot get here by driver, can the notify  pass it, like pre-version patch?

I think it's possilbe to store the voltage in driver side, and do the io-domain config after EVENT_ENABLE notify witout voltage, I will push it to io-domain maintainer to review, can I add yours' review in that patch ?
_______________________________________________
Linux-rockchip mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-rockchip