Skip to contentSkip to conversation

ARM: at91: pm: change BU Power Switch to automatic mode

PATCH Applied [email protected] 2 participants 7 messages Nov 25, 2024 – Dec 9, 2024

Subsystems (3)
lore
Share & tools
lore

[email protected]

author maintainer
#0 lore raw

PATCH 1/1 ARM: at91: pm: change BU Power Switch to automatic mode

From: Nicolas Ferre <[email protected]>

Change how the Backup Unit Power is configured and force the automatic/hardware mode.
This change eliminates the need for software management of the power switch, ensuring it transitions to the backup power source before entering low power modes.

This is done in the only locaton where this swich was configured. It's usually done in the bootloader.

Previously, the loss of the VDDANA (or VDDIN33) power source was not automatically compensated by an alternative power source. This resulted in the loss of Backup Unit content, including Backup Self-refresh low power mode information, OTP emulation configuration, and boot configuration, for instance.

---
 arch/arm/mach-at91/pm.c | 31 ++++++++++++++++++++-----------
 1 file changed, 20 insertions(+), 11 deletions(-)
arch/arm/mach-at91/pm.c +20 -11
index b9b995f8a36e..05a1547642b6 100644
@@ -598,7 +598,21 @@ static int at91_suspend_finish(unsigned long val)
 	return 0;
 }
 
-static void at91_pm_switch_ba_to_vbat(void)
+/**
+ * at91_pm_switch_ba_to_auto() - Configure Backup Unit Power Switch
+ * to automatic/hardware mode.
+ *
+ * The Backup Unit Power Switch can be managed either by software or hardware.
+ * Enabling hardware mode allows the automatic transition of power between
+ * VDDANA (or VDDIN33) and VDDBU (or VBAT, respectively), based on the
+ * availability of these power sources.
+ *
+ * If the Backup Unit Power Switch is already in automatic mode, no action is
+ * required. If it is in software-controlled mode, it is switched to automatic
+ * mode to enhance safety and eliminate the need for toggling between power
+ * sources.
+ */
+static void at91_pm_switch_ba_to_auto(void)
 {
 	unsigned int offset = offsetof(struct at91_pm_sfrbu_regs, pswbu);
 	unsigned int val;
@@ -609,24 +623,19 @@ static void at91_pm_switch_ba_to_vbat(void)
 
 	val = readl(soc_pm.data.sfrbu + offset);
 
-	/* Already on VBAT. */
-	if (!(val & soc_pm.sfrbu_regs.pswbu.state))
+	/* Already on auto/hardware. */
+	if (!(val & soc_pm.sfrbu_regs.pswbu.ctrl))
 		return;
 
-	val &= ~soc_pm.sfrbu_regs.pswbu.softsw;
-	val |= soc_pm.sfrbu_regs.pswbu.key | soc_pm.sfrbu_regs.pswbu.ctrl;
+	val &= ~soc_pm.sfrbu_regs.pswbu.ctrl;
+	val |= soc_pm.sfrbu_regs.pswbu.key;
 	writel(val, soc_pm.data.sfrbu + offset);
-
-	/* Wait for update. */
-	val = readl(soc_pm.data.sfrbu + offset);
-	while (val & soc_pm.sfrbu_regs.pswbu.state)
-		val = readl(soc_pm.data.sfrbu + offset);
 }
 
 static void at91_pm_suspend(suspend_state_t state)
 {
 	if (soc_pm.data.mode == AT91_PM_BACKUP) {
-		at91_pm_switch_ba_to_vbat();
+		at91_pm_switch_ba_to_auto();
 
 		cpu_suspend(0, at91_suspend_finish);
 
Signature · 1 line
2.39.2
maintainer
#1 lore raw

Hi, Nicolas,

Quoted 9 lines · On 25.11.2024 18:56, [email protected] wrote:

From: Nicolas Ferre <[email protected]>

Change how the Backup Unit Power is configured and force the automatic/hardware mode.
This change eliminates the need for software management of the power switch, ensuring it transitions to the backup power source before entering low power modes.

This is done in the only locaton where this swich was configured. It's

s/locaton/location

Quoted 7 lines

usually done in the bootloader.

Previously, the loss of the VDDANA (or VDDIN33) power source was not automatically compensated by an alternative power source. This resulted in the loss of Backup Unit content, including Backup Self-refresh low power mode information, OTP emulation configuration, and boot configuration, for instance.

Should we add a fixes for this?

Quoted 34 lines
---
 arch/arm/mach-at91/pm.c | 31 ++++++++++++++++++++-----------
 1 file changed, 20 insertions(+), 11 deletions(-)
arch/arm/mach-at91/pm.c +15 -1
index b9b995f8a36e..05a1547642b6 100644
@@ -598,7 +598,21 @@ static int at91_suspend_finish(unsigned long val)
 	return 0;
 }
 
-static void at91_pm_switch_ba_to_vbat(void)
+/**
+ * at91_pm_switch_ba_to_auto() - Configure Backup Unit Power Switch
+ * to automatic/hardware mode.
+ *
+ * The Backup Unit Power Switch can be managed either by software or hardware.
+ * Enabling hardware mode allows the automatic transition of power between
+ * VDDANA (or VDDIN33) and VDDBU (or VBAT, respectively), based on the
+ * availability of these power sources.
+ *
+ * If the Backup Unit Power Switch is already in automatic mode, no action is
+ * required. If it is in software-controlled mode, it is switched to automatic
+ * mode to enhance safety and eliminate the need for toggling between power
+ * sources.
+ */
+static void at91_pm_switch_ba_to_auto(void)
 {
 	unsigned int offset = offsetof(struct at91_pm_sfrbu_regs, pswbu);
 	unsigned int val;

arch/arm/mach-at91/pm.c static void at91_pm_switch_ba_to_vbat(void) · line 616

Quoted 10 lines
arch/arm/mach-at91/pm.c +2 -3
@@ -609,24 +623,19 @@ static void at91_pm_switch_ba_to_vbat(void)
 
 	val = readl(soc_pm.data.sfrbu + offset);
 
-	/* Already on VBAT. */
-	if (!(val & soc_pm.sfrbu_regs.pswbu.state))
+	/* Already on auto/hardware. */
+	if (!(val & soc_pm.sfrbu_regs.pswbu.ctrl))
 		return;
 
-	val &= ~soc_pm.sfrbu_regs.pswbu.softsw;

It seems that softsw and state members of at91_pm_sfrbu_regs.pswbu along with their initialization could be dropped. What do you think?

I can do it while applying, if any.

Thank you,
Claudiu

Quoted 19 lines
arch/arm/mach-at91/pm.c +3 -7
-	val |= soc_pm.sfrbu_regs.pswbu.key | soc_pm.sfrbu_regs.pswbu.ctrl;
+	val &= ~soc_pm.sfrbu_regs.pswbu.ctrl;
+	val |= soc_pm.sfrbu_regs.pswbu.key;
 	writel(val, soc_pm.data.sfrbu + offset);
-
-	/* Wait for update. */
-	val = readl(soc_pm.data.sfrbu + offset);
-	while (val & soc_pm.sfrbu_regs.pswbu.state)
-		val = readl(soc_pm.data.sfrbu + offset);
 }
 
 static void at91_pm_suspend(suspend_state_t state)
 {
 	if (soc_pm.data.mode == AT91_PM_BACKUP) {
-		at91_pm_switch_ba_to_vbat();
+		at91_pm_switch_ba_to_auto();
 
 		cpu_suspend(0, at91_suspend_finish);
 
author maintainer
#2 lore raw
Quoted 17 lines · Claudiu Beznea

Hi, Nicolas,

Quoted 9 lines · On 25.11.2024 18:56, [email protected] wrote:

From: Nicolas Ferre <[email protected]>

Change how the Backup Unit Power is configured and force the automatic/hardware mode.
This change eliminates the need for software management of the power switch, ensuring it transitions to the backup power source before entering low power modes.

This is done in the only locaton where this swich was configured. It's

s/locaton/location

Quoted 2 lines

usually done in the bootloader.

Quoted 7 lines
Quoted 5 lines

Previously, the loss of the VDDANA (or VDDIN33) power source was not automatically compensated by an alternative power source. This resulted in the loss of Backup Unit content, including Backup Self-refresh low power mode information, OTP emulation configuration, and boot configuration, for instance.

Should we add a fixes for this?

Not so easy to tell as there's a loose dependency with the bootloader. But it's true that switching to automatic never harm. So probably yes.

Quoted 42 lines
Quoted 42 lines
---
  arch/arm/mach-at91/pm.c | 31 ++++++++++++++++++++-----------
  1 file changed, 20 insertions(+), 11 deletions(-)
arch/arm/mach-at91/pm.c +17 -3
index b9b995f8a36e..05a1547642b6 100644
@@ -598,7 +598,21 @@ static int at91_suspend_finish(unsigned long val)
       return 0;
  }
 
-static void at91_pm_switch_ba_to_vbat(void)
+/**
+ * at91_pm_switch_ba_to_auto() - Configure Backup Unit Power Switch
+ * to automatic/hardware mode.
+ *
+ * The Backup Unit Power Switch can be managed either by software or hardware.
+ * Enabling hardware mode allows the automatic transition of power between
+ * VDDANA (or VDDIN33) and VDDBU (or VBAT, respectively), based on the
+ * availability of these power sources.
+ *
+ * If the Backup Unit Power Switch is already in automatic mode, no action is
+ * required. If it is in software-controlled mode, it is switched to automatic
+ * mode to enhance safety and eliminate the need for toggling between power
+ * sources.
+ */
+static void at91_pm_switch_ba_to_auto(void)
  {
       unsigned int offset = offsetof(struct at91_pm_sfrbu_regs, pswbu);
       unsigned int val;
@@ -609,24 +623,19 @@ static void at91_pm_switch_ba_to_vbat(void)
 
       val = readl(soc_pm.data.sfrbu + offset);
 
-     /* Already on VBAT. */
-     if (!(val & soc_pm.sfrbu_regs.pswbu.state))
+     /* Already on auto/hardware. */
+     if (!(val & soc_pm.sfrbu_regs.pswbu.ctrl))
               return;
 
Quoted 4 lines

arch/arm/mach-at91/pm.c static void at91_pm_switch_ba_to_vbat(void) · line 616

Quoted 1 line
arch/arm/mach-at91/pm.c +0 -1
-     val &= ~soc_pm.sfrbu_regs.pswbu.softsw;

It seems that softsw and state members of at91_pm_sfrbu_regs.pswbu along with their initialization could be dropped. What do you think?

I think that I tried when writing the patch but I think that there's a little difference with sama5d2 register layout. Give me a couple more days to come back to this and verify.

Quoted 25 lines

I can do it while applying, if any.

Thank you,
Claudiu

Quoted 19 lines
arch/arm/mach-at91/pm.c +3 -7
-     val |= soc_pm.sfrbu_regs.pswbu.key | soc_pm.sfrbu_regs.pswbu.ctrl;
+     val &= ~soc_pm.sfrbu_regs.pswbu.ctrl;
+     val |= soc_pm.sfrbu_regs.pswbu.key;
       writel(val, soc_pm.data.sfrbu + offset);
-
-     /* Wait for update. */
-     val = readl(soc_pm.data.sfrbu + offset);
-     while (val & soc_pm.sfrbu_regs.pswbu.state)
-             val = readl(soc_pm.data.sfrbu + offset);
  }
 
  static void at91_pm_suspend(suspend_state_t state)
  {
       if (soc_pm.data.mode == AT91_PM_BACKUP) {
-             at91_pm_switch_ba_to_vbat();
+             at91_pm_switch_ba_to_auto();
 
               cpu_suspend(0, at91_suspend_finish);
 
author maintainer
#3 lore raw

Claudiu,

Quoted 71 lines · Nicolas Ferre
Quoted 17 lines · Claudiu Beznea

Hi, Nicolas,

Quoted 9 lines · On 25.11.2024 18:56, [email protected] wrote:

From: Nicolas Ferre <[email protected]>

Change how the Backup Unit Power is configured and force the automatic/hardware mode.
This change eliminates the need for software management of the power switch, ensuring it transitions to the backup power source before entering low power modes.

This is done in the only locaton where this swich was configured. It's

s/locaton/location

Quoted 2 lines

usually done in the bootloader.

Quoted 7 lines
Quoted 5 lines

Previously, the loss of the VDDANA (or VDDIN33) power source was not automatically compensated by an alternative power source. This resulted in the loss of Backup Unit content, including Backup Self-refresh low power mode information, OTP emulation configuration, and boot configuration, for instance.

Should we add a fixes for this?

Not so easy to tell as there's a loose dependency with the bootloader. But it's true that switching to automatic never harm. So probably yes.

Quoted 42 lines
Quoted 42 lines
---
   arch/arm/mach-at91/pm.c | 31 ++++++++++++++++++++-----------
   1 file changed, 20 insertions(+), 11 deletions(-)
arch/arm/mach-at91/pm.c +17 -3
index b9b995f8a36e..05a1547642b6 100644
@@ -598,7 +598,21 @@ static int at91_suspend_finish(unsigned long val)
        return 0;
   }
 
-static void at91_pm_switch_ba_to_vbat(void)
+/**
+ * at91_pm_switch_ba_to_auto() - Configure Backup Unit Power Switch
+ * to automatic/hardware mode.
+ *
+ * The Backup Unit Power Switch can be managed either by software or hardware.
+ * Enabling hardware mode allows the automatic transition of power between
+ * VDDANA (or VDDIN33) and VDDBU (or VBAT, respectively), based on the
+ * availability of these power sources.
+ *
+ * If the Backup Unit Power Switch is already in automatic mode, no action is
+ * required. If it is in software-controlled mode, it is switched to automatic
+ * mode to enhance safety and eliminate the need for toggling between power
+ * sources.
+ */
+static void at91_pm_switch_ba_to_auto(void)
   {
        unsigned int offset = offsetof(struct at91_pm_sfrbu_regs, pswbu);
        unsigned int val;
@@ -609,24 +623,19 @@ static void at91_pm_switch_ba_to_vbat(void)
 
        val = readl(soc_pm.data.sfrbu + offset);
 
-     /* Already on VBAT. */
-     if (!(val & soc_pm.sfrbu_regs.pswbu.state))
+     /* Already on auto/hardware. */
+     if (!(val & soc_pm.sfrbu_regs.pswbu.ctrl))
                return;
 
Quoted 8 lines
Quoted 4 lines

arch/arm/mach-at91/pm.c static void at91_pm_switch_ba_to_vbat(void) · line 616

Quoted 1 line
arch/arm/mach-at91/pm.c +0 -1
-     val &= ~soc_pm.sfrbu_regs.pswbu.softsw;

It seems that softsw and state members of at91_pm_sfrbu_regs.pswbu along with their initialization could be dropped. What do you think?

I think that I tried when writing the patch but I think that there's a little difference with sama5d2 register layout. Give me a couple more days to come back to this and verify.

Ok, I remember now: I was wondering if I needed to remove the whole sfrbu_regs.xxx mechanism and define more generically the content of include/soc/at91/sama7-sfrbu.h for sama5d2, but if we need one day to use the STATE bit or even the SMCTRL bit of sama5d2, then it should be kept.

So, now that the mechanism is in place, I would prefer that we keep it: okay for you?

Do you want me to re-spin a v2 for the rest?

Best regards,
Nicolas

Quoted 27 lines
Quoted 25 lines

I can do it while applying, if any.

Thank you,
Claudiu

Quoted 19 lines
arch/arm/mach-at91/pm.c +3 -7
-     val |= soc_pm.sfrbu_regs.pswbu.key | soc_pm.sfrbu_regs.pswbu.ctrl;
+     val &= ~soc_pm.sfrbu_regs.pswbu.ctrl;
+     val |= soc_pm.sfrbu_regs.pswbu.key;
        writel(val, soc_pm.data.sfrbu + offset);
-
-     /* Wait for update. */
-     val = readl(soc_pm.data.sfrbu + offset);
-     while (val & soc_pm.sfrbu_regs.pswbu.state)
-             val = readl(soc_pm.data.sfrbu + offset);
   }
 
   static void at91_pm_suspend(suspend_state_t state)
   {
        if (soc_pm.data.mode == AT91_PM_BACKUP) {
-             at91_pm_switch_ba_to_vbat();
+             at91_pm_switch_ba_to_auto();
 
                cpu_suspend(0, at91_suspend_finish);
 
maintainer
#4 lore raw

Hi, Nicolas,

Quoted 88 lines · Nicolas Ferre

Claudiu,

Quoted 76 lines · Nicolas Ferre
Quoted 17 lines · Claudiu Beznea

Hi, Nicolas,

Quoted 9 lines · On 25.11.2024 18:56, [email protected] wrote:

From: Nicolas Ferre <[email protected]>

Change how the Backup Unit Power is configured and force the automatic/hardware mode.
This change eliminates the need for software management of the power switch, ensuring it transitions to the backup power source before entering low power modes.

This is done in the only locaton where this swich was configured. It's

s/locaton/location

Quoted 2 lines

usually done in the bootloader.

Quoted 7 lines
Quoted 5 lines

Previously, the loss of the VDDANA (or VDDIN33) power source was not automatically compensated by an alternative power source. This resulted in the loss of Backup Unit content, including Backup Self-refresh low power mode information, OTP emulation configuration, and boot configuration, for instance.

Should we add a fixes for this?

Not so easy to tell as there's a loose dependency with the bootloader. But it's true that switching to automatic never harm. So probably yes.

Quoted 47 lines
Quoted 47 lines

---
   arch/arm/mach-at91/pm.c | 31 ++++++++++++++++++++-----------

   1 file changed, 20 insertions(+), 11 deletions(-)
arch/arm/mach-at91/pm.c +5 -1
index b9b995f8a36e..05a1547642b6 100644
@@ -598,7 +598,21 @@ static int at91_suspend_finish(unsigned long val)
         return 0;
    }
 
-static void at91_pm_switch_ba_to_vbat(void)
+/**
+ * at91_pm_switch_ba_to_auto() - Configure Backup Unit Power Switch
+ * to automatic/hardware mode.
+ *
+ * The Backup Unit Power Switch can be managed either by software or

hardware.
+ * Enabling hardware mode allows the automatic transition of power between
+ * VDDANA (or VDDIN33) and VDDBU (or VBAT, respectively), based on the
+ * availability of these power sources.
+ *
+ * If the Backup Unit Power Switch is already in automatic mode, no action is
+ * required. If it is in software-controlled mode, it is switched to automatic
+ * mode to enhance safety and eliminate the need for toggling between power
+ * sources.
+ */
+static void at91_pm_switch_ba_to_auto(void)
   {
        unsigned int offset = offsetof(struct at91_pm_sfrbu_regs, pswbu); unsigned int val;

arch/arm/mach-at91/pm.c +2 -2
@@ -609,24 +623,19 @@ static void at91_pm_switch_ba_to_vbat(void)
 
         val = readl(soc_pm.data.sfrbu + offset);
 
-     /* Already on VBAT. */
-     if (!(val & soc_pm.sfrbu_regs.pswbu.state))
+     /* Already on auto/hardware. */
+     if (!(val & soc_pm.sfrbu_regs.pswbu.ctrl))
                 return;
 
Quoted 8 lines
Quoted 4 lines

arch/arm/mach-at91/pm.c static void at91_pm_switch_ba_to_vbat(void) · line 616

Quoted 1 line
arch/arm/mach-at91/pm.c +0 -1
-     val &= ~soc_pm.sfrbu_regs.pswbu.softsw;

It seems that softsw and state members of at91_pm_sfrbu_regs.pswbu along with their initialization could be dropped. What do you think?

I think that I tried when writing the patch but I think that there's a little difference with sama5d2 register layout. Give me a couple more days to come back to this and verify.

Quoted 7 lines

Ok, I remember now: I was wondering if I needed to remove the whole sfrbu_regs.xxx mechanism and define more generically the content of include/soc/at91/sama7-sfrbu.h for sama5d2, but if we need one day to use the STATE bit or even the SMCTRL bit of sama5d2, then it should be kept.

So, now that the mechanism is in place, I would prefer that we keep it:
okay for you?

OK

Quoted 2 lines

Do you want me to re-spin a v2 for the rest?

No, I can handle the insignificant typo while applying.

Thank you,
Claudiu

Quoted 32 lines

Best regards,
  Nicolas

Quoted 27 lines
Quoted 25 lines

I can do it while applying, if any.

Thank you,
Claudiu

Quoted 19 lines
arch/arm/mach-at91/pm.c +3 -7
-     val |= soc_pm.sfrbu_regs.pswbu.key | soc_pm.sfrbu_regs.pswbu.ctrl;
+     val &= ~soc_pm.sfrbu_regs.pswbu.ctrl;
+     val |= soc_pm.sfrbu_regs.pswbu.key;
         writel(val, soc_pm.data.sfrbu + offset);
-
-     /* Wait for update. */
-     val = readl(soc_pm.data.sfrbu + offset);
-     while (val & soc_pm.sfrbu_regs.pswbu.state)
-             val = readl(soc_pm.data.sfrbu + offset);
    }
 
    static void at91_pm_suspend(suspend_state_t state)
    {
         if (soc_pm.data.mode == AT91_PM_BACKUP) {
-             at91_pm_switch_ba_to_vbat();
+             at91_pm_switch_ba_to_auto();
 
                 cpu_suspend(0, at91_suspend_finish);
 
maintainer Applied
#5 lore raw
In reply to [email protected]
Quoted 11 lines · On 25.11.2024 18:56, [email protected] wrote:

From: Nicolas Ferre <[email protected]>

Change how the Backup Unit Power is configured and force the automatic/hardware mode.
This change eliminates the need for software management of the power switch, ensuring it transitions to the backup power source before entering low power modes.

This is done in the only locaton where this swich was configured. It's usually done in the bootloader.

Quoted 7 lines

Previously, the loss of the VDDANA (or VDDIN33) power source was not automatically compensated by an alternative power source. This resulted in the loss of Backup Unit content, including Backup Self-refresh low power mode information, OTP emulation configuration, and boot configuration, for instance.

I've added the fixes tag, s/locaton/location, s/swich/switch.

Applied to at91-soc, thanks!

author maintainer
#6 lore raw
Quoted 18 lines · Claudiu Beznea
Quoted 17 lines · On 25.11.2024 18:56, [email protected] wrote:

From: Nicolas Ferre <[email protected]>

Change how the Backup Unit Power is configured and force the automatic/hardware mode.
This change eliminates the need for software management of the power switch, ensuring it transitions to the backup power source before entering low power modes.

This is done in the only locaton where this swich was configured. It's usually done in the bootloader.

Previously, the loss of the VDDANA (or VDDIN33) power source was not automatically compensated by an alternative power source. This resulted in the loss of Backup Unit content, including Backup Self-refresh low power mode information, OTP emulation configuration, and boot configuration, for instance.

Quoted 3 lines
Quoted 1 line

I've added the fixes tag, s/locaton/location, s/swich/switch.

Thanks for your help Claudiu. Regards,
Nicolas

Quoted 2 lines

Applied to at91-soc, thanks!