Skip to contentSkip to conversation

starfive reboot cleanup & fix

PATCH New Nam Cao 3 participants 10 messages Jan 31, 2024 – Feb 2, 2024

lore
Share & tools
lore

Nam Cao

author
#0 lore raw

cover starfive reboot cleanup & fix

This series:
- add gracefully error handling when reboot devices are not present
- correct a device tree compatibility string
- some minor cleanup

This series complements my other series:
http://lists.infradead.org/pipermail/opensbi/2024-January/006300.html

Nam Cao (5):
platform: starfive: remove redundant compatibility check in pmic_ops
platform: starfive: rename "stf,axp15060-regulator" ->

    "x-powers,axp15060"
  platform: starfive: drop "compatible" from struct pmic
  platform: starfive: return error if needed devices are not present
  platform: starfive: call starfive_jh7110_inst_init() in
    pm_reset_init()
 platform/generic/starfive/jh7110.c | 49 ++++++++++++++++--------------
 1 file changed, 27 insertions(+), 22 deletions(-)
Signature · 1 line
2.39.2
author
#1 lore raw

PATCH 1/5 platform: starfive: remove redundant compatibility check in pmic_ops

pmic_ops() is only called if a compatible device is found in device tree. It is redundant for this function to check the compability again. Remove this check.

---
 platform/generic/starfive/jh7110.c | 36 ++++++++++++++----------------
 1 file changed, 17 insertions(+), 19 deletions(-)
platform/generic/starfive/jh7110.c +17 -19
index dcd6306..72969e4 100644
@@ -134,28 +134,26 @@ static void pmic_ops(struct pmic *pmic, int type)
 	if (ret)
 		return;
 
-	if (!sbi_strcmp("stf,axp15060-regulator", pmic->compatible)) {
-		ret = i2c_adapter_reg_read(pmic->adapter, pmic->dev_addr,
-					   AXP15060_POWER_REG, &val);
-
-		if (ret) {
-			sbi_printf("%s: cannot read pmic power register\n",
-				   __func__);
-			return;
-		}
+	ret = i2c_adapter_reg_read(pmic->adapter, pmic->dev_addr,
+					AXP15060_POWER_REG, &val);
+
+	if (ret) {
+		sbi_printf("%s: cannot read pmic power register\n",
+				__func__);
+		return;
+	}
 
+	val |= AXP15060_POWER_OFF_BIT;
+	if (type == SBI_SRST_RESET_TYPE_SHUTDOWN)
 		val |= AXP15060_POWER_OFF_BIT;
-		if (type == SBI_SRST_RESET_TYPE_SHUTDOWN)
-			val |= AXP15060_POWER_OFF_BIT;
-		else
-			val |= AXP15060_RESET_BIT;
+	else
+		val |= AXP15060_RESET_BIT;
 
-		ret = i2c_adapter_reg_write(pmic->adapter, pmic->dev_addr,
-					    AXP15060_POWER_REG, val);
-		if (ret)
-			sbi_printf("%s: cannot write pmic power register\n",
-				   __func__);
-	}
+	ret = i2c_adapter_reg_write(pmic->adapter, pmic->dev_addr,
+					AXP15060_POWER_REG, val);
+	if (ret)
+		sbi_printf("%s: cannot write pmic power register\n",
+				__func__);
 }
 
 static void pmic_i2c_clk_enable(void)
Signature · 1 line
2.39.2
author
#2 lore raw

PATCH 2/5 platform: starfive: rename "stf,axp15060-regulator" -> "x-powers,axp15060"

In reply to Nam Cao

OpenSBI uses the device tree compatible string "stf,axp15060-regulator" for the regulator node. However, the string used by U-Boot (and Linux) is actually "x-powers,axp15060". As OpenSBI gets the device tree from U-Boot, this causes the regulator device to be undetected, and OpenSBI does not use this device to perform board reset/shutdown.

Rename this device tree compatible string to match U-Boot (and Linux).

---
Actually "x-powers,axp15060" does not exist in U-Boot at the moment. But
it is being added:
https://lore.kernel.org/u-boot/47fd2b2225f52a2d5c3a91494ef7b9036e944199.1706517351.git.namcao at linutronix.de/

 platform/generic/starfive/jh7110.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
platform/generic/starfive/jh7110.c +1 -1
index 72969e4..9f5261b 100644
@@ -227,7 +227,7 @@ static int pm_reset_init(void *fdt, int nodeoff,
 }
 
 static const struct fdt_match pm_reset_match[] = {
-	{ .compatible = "stf,axp15060-regulator", .data = (void *)true },
+	{ .compatible = "x-powers,axp15060", .data = (void *)true },
 	{ },
 };
 
Signature · 1 line
2.39.2
author
#3 lore raw

PATCH 3/5 platform: starfive: drop "compatible" from struct pmic

In reply to Nam Cao

The member "compatible" of struct pmic is not used. Delete it.

---
 platform/generic/starfive/jh7110.c | 2 --
 1 file changed, 2 deletions(-)
platform/generic/starfive/jh7110.c +0 -2
index 9f5261b..cb8c48d 100644
@@ -23,7 +23,6 @@
 struct pmic {
 	struct i2c_adapter *adapter;
 	u32 dev_addr;
-	const char *compatible;
 };
 
 struct jh7110 {
@@ -208,7 +207,6 @@ static int pm_reset_init(void *fdt, int nodeoff,
 		return rc;
 
 	pmic_inst.dev_addr = addr;
-	pmic_inst.compatible = match->compatible;
 
 	i2c_bus = fdt_parent_offset(fdt, nodeoff);
 	if (i2c_bus < 0)
Signature · 1 line
2.39.2
author
#4 lore raw

PATCH 4/5 platform: starfive: return error if needed devices are not present

In reply to Nam Cao

Jh7110's reset driver needs power management device and clock controller device to work. Currently, the driver proceed anyway without these devices, and invalid addresses (jh7110_inst.pmu_reg_base and jh7110_inst.clk_reg_base) are used during reboot, which causes unpredictable broken behaviors.

If these devices are not present, return -SBI_ENODEV.

---
 platform/generic/starfive/jh7110.c | 4 ++++
 1 file changed, 4 insertions(+)
platform/generic/starfive/jh7110.c +4 -0
index cb8c48d..f9942d3 100644
@@ -246,6 +246,8 @@ static int starfive_jh7110_inst_init(void *fdt)
 		if (rc)
 			goto err;
 		jh7110_inst.pmu_reg_base = addr;
+	} else {
+		return -SBI_ENODEV;
 	}
 
 	noff = fdt_node_offset_by_compatible(fdt, -1, "starfive,jh7110-clkgen");
@@ -254,6 +256,8 @@ static int starfive_jh7110_inst_init(void *fdt)
 		if (rc)
 			goto err;
 		jh7110_inst.clk_reg_base = addr;
+	} else {
+		return -SBI_ENODEV;
 	}
 
 	if (pmic_inst.adapter) {
Signature · 1 line
2.39.2
author
#5 lore raw

PATCH 5/5 platform: starfive: call starfive_jh7110_inst_init() in pm_reset_init()

In reply to Nam Cao

The function starfive_jh7110_inst_init() initialize some power management unit address and clock addresses, needed for the reset driver. It doesn't do anything else, and also the reset driver doesn't work without calling this function. Thus, it does not make much sense that this function is independent from pm_reset_init().

Delete the separate call to starfive_jh7110_inst_init(), and instead just call this function inside pm_reset_init().

Doing this also fixes another problem: if starfive_jh7110_inst_init() returns an error code, it gets propagated to final_init() and OpenSBI hangs. This hang is not necessary, because failures within starfive_jh7110_inst_init() only mean OpenSBI cannot perform reboot or shutdown, but the system can still function normally.

---
 platform/generic/starfive/jh7110.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)
platform/generic/starfive/jh7110.c +6 -1
index f9942d3..520bc8a 100644
@@ -194,6 +194,8 @@ static struct sbi_system_reset_device pm_reset = {
 	.system_reset = pm_system_reset
 };
 
+static int starfive_jh7110_inst_init(void *fdt);
+
 static int pm_reset_init(void *fdt, int nodeoff,
 			 const struct fdt_match *match)
 {
@@ -219,6 +221,10 @@ static int pm_reset_init(void *fdt, int nodeoff,
 
 	pmic_inst.adapter = adapter;
 
+	rc = starfive_jh7110_inst_init(fdt);
+	if (rc)
+		return rc;
+
 	sbi_system_reset_add_device(&pm_reset);
 
 	return 0;
@@ -278,7 +284,6 @@ static int starfive_jh7110_final_init(bool cold_boot,
 
 	if (cold_boot) {
 		fdt_reset_driver_init(fdt, &fdt_reset_pmic);
-		return starfive_jh7110_inst_init(fdt);
 	}
 
 	return 0;
Signature · 1 line
2.39.2
#6 lore raw

cover starfive reboot cleanup & fix

In reply to Nam Cao
Quoted 9 lines · On Wed, Jan 31, 2024 at 6:39?PM Nam Cao <[email protected]> wrote:

This series:
- add gracefully error handling when reboot devices are not present
- correct a device tree compatibility string
- some minor cleanup

This series complements my other series:
http://lists.infradead.org/pipermail/opensbi/2024-January/006300.html

Quoted 8 lines

Nam Cao (5):
platform: starfive: remove redundant compatibility check in pmic_ops
platform: starfive: rename "stf,axp15060-regulator" ->

    "x-powers,axp15060"
  platform: starfive: drop "compatible" from struct pmic
  platform: starfive: return error if needed devices are not present
  platform: starfive: call starfive_jh7110_inst_init() in
    pm_reset_init()

Can someone from StarFive review and test this series ?

Regards,
Anup

Quoted 12 lines
 platform/generic/starfive/jh7110.c | 49 ++++++++++++++++--------------
 1 file changed, 27 insertions(+), 22 deletions(-)

--
2.39.2

--
opensbi mailing list
opensbi at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/opensbi

#7 lore raw

cover starfive reboot cleanup & fix

Quoted 21 lines · On Wed, Jan 31, 2024 at 6:39?PM Nam Cao <[email protected]> wrote:
Quoted 9 lines

This series:
- add gracefully error handling when reboot devices are not present
- correct a device tree compatibility string
- some minor cleanup

This series complements my other series:
http://lists.infradead.org/pipermail/opensbi/2024-January/006300.html

Quoted 8 lines

Nam Cao (5):
platform: starfive: remove redundant compatibility check in pmic_ops
platform: starfive: rename "stf,axp15060-regulator" ->

    "x-powers,axp15060"
  platform: starfive: drop "compatible" from struct pmic
  platform: starfive: return error if needed devices are not present
  platform: starfive: call starfive_jh7110_inst_init() in
    pm_reset_init()
Quoted 5 lines

Can someone from StarFive review and test this series ?

Regards,
Anup

Let me review these. But I don't receive this series patch.

Minda

Quoted 12 lines
Quoted 12 lines
 platform/generic/starfive/jh7110.c | 49 ++++++++++++++++--------------
 1 file changed, 27 insertions(+), 22 deletions(-)

--
2.39.2

--
opensbi mailing list
opensbi at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/opensbi

author
#8 lore raw

cover starfive reboot cleanup & fix

Quoted 19 lines · On Thu, 1 Feb 2024 05:38:16 +0000 Minda Chen <[email protected]>…
Quoted 19 lines · On Wed, Jan 31, 2024 at 6:39?PM Nam Cao <[email protected]> wrote:
Quoted 17 lines

This series:
- add gracefully error handling when reboot devices are not present
- correct a device tree compatibility string
- some minor cleanup

This series complements my other series:
http://lists.infradead.org/pipermail/opensbi/2024-January/006300.html

Nam Cao (5):
platform: starfive: remove redundant compatibility check in pmic_ops
platform: starfive: rename "stf,axp15060-regulator" ->

    "x-powers,axp15060"
  platform: starfive: drop "compatible" from struct pmic
  platform: starfive: return error if needed devices are not present
  platform: starfive: call starfive_jh7110_inst_init() in
    pm_reset_init()  
Quoted 6 lines
Quoted 5 lines

Can someone from StarFive review and test this series ?

Regards,
Anup

Let me review these. But I don't receive this series patch.

I only sent it to the list. Can you grab the series from there?

Or I can also resend the series and include you as a recipient, if that's more convenient.

Best regards,
Nam

#9 lore raw

cover starfive reboot cleanup & fix

Quoted 32 lines
Quoted 20 lines · On Thu, 1 Feb 2024 05:38:16 +0000 Minda Chen <[email protected]>…
Quoted 20 lines · On Wed, Jan 31, 2024 at 6:39?PM Nam Cao <[email protected]> wrote:
Quoted 18 lines

This series:
- add gracefully error handling when reboot devices are not present
- correct a device tree compatibility string
- some minor cleanup

This series complements my other series:
http://lists.infradead.org/pipermail/opensbi/2024-January/006300.h
tml

Nam Cao (5):
platform: starfive: remove redundant compatibility check in pmic_ops
platform: starfive: rename "stf,axp15060-regulator" ->

    "x-powers,axp15060"
  platform: starfive: drop "compatible" from struct pmic
  platform: starfive: return error if needed devices are not present
  platform: starfive: call starfive_jh7110_inst_init() in
    pm_reset_init()
Quoted 6 lines
Quoted 5 lines

Can someone from StarFive review and test this series ?

Regards,
Anup

Let me review these. But I don't receive this series patch.

Quoted 7 lines

I only sent it to the list. Can you grab the series from there?

Or I can also resend the series and include you as a recipient, if that's more convenient.

Best regards,
Nam

Resend the patchset is better. Thanks
Minda