Skip to contentSkip to conversation

pinctrl: amd: debounce filter fixes

PATCH In review Coiby Xu 4 participants 15 messages Nov 5, 2020 – Nov 10, 2020

Subsystems (4)
lore
Share & tools
lore

Coiby Xu

author
#0 lore raw

v3 cover pinctrl: amd: debounce filter fixes

There are three fixes (and one improvement) that fix the debounce filter issues in pinctrl-amd.

Changelog v3:
- Explain why the driver mistakenly took the slightly deviated value of debounce timeout in the commit message (patch 2/4) [Andy Shevchenko]
- Explain why other touchpads are affected by the issue of setting debounce filter in IRQ type setting in the commit message (patch 4/4)

Changelog v2:
- Message-Id to Link and grammar fixes for commit messages [Andy Shevchenko]

Coiby Xu (4):
pinctrl: amd: fix incorrect way to disable debounce filter
pinctrl: amd: use higher precision for 512 RtcClk
pinctrl: amd: print debounce filter info in debugfs
pinctrl: amd: remove debounce filter setting in IRQ type setting

 drivers/pinctrl/pinctrl-amd.c | 56 +++++++++++++++++++++++++++--------
 1 file changed, 44 insertions(+), 12 deletions(-)

--
2.28.0

author
#1 lore raw

PATCH v3 1/4 pinctrl: amd: fix incorrect way to disable debounce filter

The correct way to disable debounce filter is to clear bit 5 and 6 of the register.

---
 drivers/pinctrl/pinctrl-amd.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/pinctrl/pinctrl-amd.c +2 -2
index 9a760f5cd7ed..d6b2b4bd337c 100644
@@ -166,14 +166,14 @@ static int amd_gpio_set_debounce(struct gpio_chip *gc, unsigned offset,
 			pin_reg |= BIT(DB_TMR_OUT_UNIT_OFF);
 			pin_reg |= BIT(DB_TMR_LARGE_OFF);
 		} else {
-			pin_reg &= ~DB_CNTRl_MASK;
+			pin_reg &= ~(DB_CNTRl_MASK << DB_CNTRL_OFF);
 			ret = -EINVAL;
 		}
 	} else {
 		pin_reg &= ~BIT(DB_TMR_OUT_UNIT_OFF);
 		pin_reg &= ~BIT(DB_TMR_LARGE_OFF);
 		pin_reg &= ~DB_TMR_OUT_MASK;
-		pin_reg &= ~DB_CNTRl_MASK;
+		pin_reg &= ~(DB_CNTRl_MASK << DB_CNTRL_OFF);
 	}
 	writel(pin_reg, gpio_dev->base + offset * 4);
 	raw_spin_unlock_irqrestore(&gpio_dev->lock, flags);
Signature · 1 line
2.28.0
Reviewed-by on v3 1/4
#2 lore raw

Re: [PATCH v3 1/4] pinctrl: amd: fix incorrect way to disable debounce filter

Hi,

Quoted 6 lines · Coiby Xu

The correct way to disable debounce filter is to clear bit 5 and 6 of the register.

Thanks, patch looks good to me:

Regards,

Hans

Quoted 26 lines
---
 drivers/pinctrl/pinctrl-amd.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/pinctrl/pinctrl-amd.c +2 -2
index 9a760f5cd7ed..d6b2b4bd337c 100644
@@ -166,14 +166,14 @@ static int amd_gpio_set_debounce(struct gpio_chip *gc, unsigned offset,
 			pin_reg |= BIT(DB_TMR_OUT_UNIT_OFF);
 			pin_reg |= BIT(DB_TMR_LARGE_OFF);
 		} else {
-			pin_reg &= ~DB_CNTRl_MASK;
+			pin_reg &= ~(DB_CNTRl_MASK << DB_CNTRL_OFF);
 			ret = -EINVAL;
 		}
 	} else {
 		pin_reg &= ~BIT(DB_TMR_OUT_UNIT_OFF);
 		pin_reg &= ~BIT(DB_TMR_LARGE_OFF);
 		pin_reg &= ~DB_TMR_OUT_MASK;
-		pin_reg &= ~DB_CNTRl_MASK;
+		pin_reg &= ~(DB_CNTRl_MASK << DB_CNTRL_OFF);
 	}
 	writel(pin_reg, gpio_dev->base + offset * 4);
 	raw_spin_unlock_irqrestore(&gpio_dev->lock, flags);
author
#4 lore raw

PATCH v3 2/4 pinctrl: amd: use higher precision for 512 RtcClk

In reply to Coiby Xu

RTC is 32.768kHz thus 512 RtcClk equals 15625 usec. The documentation likely has dropped precision and that's why the driver mistakenly took the slightly deviated value.

---
 drivers/pinctrl/pinctrl-amd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/pinctrl/pinctrl-amd.c +1 -1
index d6b2b4bd337c..4aea3e05e8c6 100644
@@ -156,7 +156,7 @@ static int amd_gpio_set_debounce(struct gpio_chip *gc, unsigned offset,
 			pin_reg |= BIT(DB_TMR_OUT_UNIT_OFF);
 			pin_reg &= ~BIT(DB_TMR_LARGE_OFF);
 		} else if (debounce < 250000) {
-			time = debounce / 15600;
+			time = debounce / 15625;
 			pin_reg |= time & DB_TMR_OUT_MASK;
 			pin_reg &= ~BIT(DB_TMR_OUT_UNIT_OFF);
 			pin_reg |= BIT(DB_TMR_LARGE_OFF);
Signature · 1 line
2.28.0
Reviewed-by on v3 2/4
#5 lore raw

Re: [PATCH v3 2/4] pinctrl: amd: use higher precision for 512 RtcClk

Hi,

Quoted 10 lines · Coiby Xu

RTC is 32.768kHz thus 512 RtcClk equals 15625 usec. The documentation likely has dropped precision and that's why the driver mistakenly took the slightly deviated value.

Thanks, patch looks good to me:

Regards,

Hans

Quoted 18 lines
---
 drivers/pinctrl/pinctrl-amd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/pinctrl/pinctrl-amd.c +1 -1
index d6b2b4bd337c..4aea3e05e8c6 100644
@@ -156,7 +156,7 @@ static int amd_gpio_set_debounce(struct gpio_chip *gc, unsigned offset,
 			pin_reg |= BIT(DB_TMR_OUT_UNIT_OFF);
 			pin_reg &= ~BIT(DB_TMR_LARGE_OFF);
 		} else if (debounce < 250000) {
-			time = debounce / 15600;
+			time = debounce / 15625;
 			pin_reg |= time & DB_TMR_OUT_MASK;
 			pin_reg &= ~BIT(DB_TMR_OUT_UNIT_OFF);
 			pin_reg |= BIT(DB_TMR_LARGE_OFF);
#6 lore raw

Re: [PATCH v3 2/4] pinctrl: amd: use higher precision for 512 RtcClk

In reply to Coiby Xu
Quoted 10 lines · Coiby Xu

RTC is 32.768kHz thus 512 RtcClk equals 15625 usec. The documentation likely has dropped precision and that's why the driver mistakenly took the slightly deviated value.

This patch applied for fixes and tagged for stable.

Yours,
Linus Walleij

author
#7 lore raw

PATCH v3 3/4 pinctrl: amd: print debounce filter info in debugfs

In reply to Coiby Xu

Print the status of debounce filter as follows,

$ cat /sys/kernel/debug/gpio
pin129          interrupt is disabled| interrupt is masked| disable wakeup in S0i3 state| disable wakeup in S3 state|
 disable wakeup in S4/S5 state| input is high|   pull-up is disabled| Pull-down is disabled|   output is disabled| debouncing filter disabled|   0x50000
                                                                                                                   ^^^^^^^^^^^^^^^^^^^^^^^^^^
pin130          interrupt is disabled| interrupt is masked| disable wakeup in S0i3 state| disable wakeup in S3 state|
 disable wakeup in S4/S5 state| input is high|   pull-up is disabled| Pull-down is disabled|   output is disabled| debouncing filter (high) enabled| debouncing timeout is 124800 (us)| 0x503c8
                                                                                                                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
---
 drivers/pinctrl/pinctrl-amd.c | 43 +++++++++++++++++++++++++++++++++--
 1 file changed, 41 insertions(+), 2 deletions(-)
drivers/pinctrl/pinctrl-amd.c +41 -2
index 4aea3e05e8c6..e9b761c2b77a 100644
@@ -197,10 +197,16 @@ static int amd_gpio_set_config(struct gpio_chip *gc, unsigned offset,
 static void amd_gpio_dbg_show(struct seq_file *s, struct gpio_chip *gc)
 {
 	u32 pin_reg;
+	u32 db_cntrl;
 	unsigned long flags;
 	unsigned int bank, i, pin_num;
 	struct amd_gpio *gpio_dev = gpiochip_get_data(gc);
 
+	bool tmr_out_unit;
+	unsigned int time;
+	unsigned int unit;
+	bool tmr_large;
+
 	char *level_trig;
 	char *active_level;
 	char *interrupt_enable;
@@ -214,6 +220,8 @@ static void amd_gpio_dbg_show(struct seq_file *s, struct gpio_chip *gc)
 	char *pull_down_enable;
 	char *output_value;
 	char *output_enable;
+	char debounce_value[40];
+	char *debounce_enable;
 
 	for (bank = 0; bank < gpio_dev->hwbank_num; bank++) {
 		seq_printf(s, "GPIO bank%d\t", bank);
@@ -327,13 +335,44 @@ static void amd_gpio_dbg_show(struct seq_file *s, struct gpio_chip *gc)
 					pin_sts = "input is low|";
 			}
 
+			db_cntrl = (DB_CNTRl_MASK << DB_CNTRL_OFF) & pin_reg;
+			if (db_cntrl) {
+				tmr_out_unit = pin_reg & BIT(DB_TMR_OUT_UNIT_OFF);
+				tmr_large = pin_reg & BIT(DB_TMR_LARGE_OFF);
+				time = pin_reg & DB_TMR_OUT_MASK;
+				if (tmr_large) {
+					if (tmr_out_unit)
+						unit = 62500;
+					else
+						unit = 15625;
+				} else {
+					if (tmr_out_unit)
+						unit = 244;
+					else
+						unit = 61;
+				}
+				if ((DB_TYPE_REMOVE_GLITCH << DB_CNTRL_OFF) == db_cntrl)
+					debounce_enable = "debouncing filter (high and low) enabled|";
+				else if ((DB_TYPE_PRESERVE_LOW_GLITCH << DB_CNTRL_OFF) == db_cntrl)
+					debounce_enable = "debouncing filter (low) enabled|";
+				else
+					debounce_enable = "debouncing filter (high) enabled|";
+
+				snprintf(debounce_value, sizeof(debounce_value),
+					 "debouncing timeout is %u (us)|", time * unit);
+			} else {
+				debounce_enable = "debouncing filter disabled|";
+				snprintf(debounce_value, sizeof(debounce_value), " ");
+			}
+
 			seq_printf(s, "%s %s %s %s %s %s\n"
-				" %s %s %s %s %s %s %s 0x%x\n",
+				" %s %s %s %s %s %s %s %s %s 0x%x\n",
 				level_trig, active_level, interrupt_enable,
 				interrupt_mask, wake_cntrl0, wake_cntrl1,
 				wake_cntrl2, pin_sts, pull_up_sel,
 				pull_up_enable, pull_down_enable,
-				output_value, output_enable, pin_reg);
+				output_value, output_enable,
+				debounce_enable, debounce_value, pin_reg);
 		}
 	}
 }
Signature · 1 line
2.28.0
#8 lore raw

Re: [PATCH v3 3/4] pinctrl: amd: print debounce filter info in debugfs

Quoted 10 lines · Coiby Xu

Print the status of debounce filter as follows,

$ cat /sys/kernel/debug/gpio
pin129          interrupt is disabled| interrupt is masked| disable wakeup in S0i3 state| disable wakeup in S3 state|
 disable wakeup in S4/S5 state| input is high|   pull-up is disabled| Pull-down is disabled|   output is disabled| debouncing filter disabled|   0x50000
                                                                                                                   ^^^^^^^^^^^^^^^^^^^^^^^^^^
pin130          interrupt is disabled| interrupt is masked| disable wakeup in S0i3 state| disable wakeup in S3 state|
 disable wakeup in S4/S5 state| input is high|   pull-up is disabled| Pull-down is disabled|   output is disabled| debouncing filter (high) enabled| debouncing timeout is 124800 (us)| 0x503c8
                                                                                                                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Patch applied for the next kernel (v5.11).

Yours,
Linus Walleij

author
#9 lore raw

PATCH v3 4/4 pinctrl: amd: remove debounce filter setting in IRQ type setting

In reply to Coiby Xu

Debounce filter setting should be independent from IRQ type setting because according to the ACPI specs, there are separate arguments for specifying debounce timeout and IRQ type in GpioIo() and GpioInt().

This will fix broken touchpads for laptops whose BIOS set the debounce timeout to a relatively large value. For example, the BIOS of Lenovo Legion-5 AMD gaming laptops including 15ARH05 (R7000) and R7000P set the debounce timeout to 124.8ms. This led to the kernel receiving only ~7 HID reports per second from the Synaptics touchpad (MSFT0001:00 06CB:7F28). Existing touchpads like [1][2] are not troubled by this bug because the debounce timeout has been set to 0 by the BIOS before enabling the debounce filter in setting IRQ type.

[1] https://github.com/Syniurge/i2c-amd-mp2/issues/11#issuecomment-721331582
[2] https://forum.manjaro.org/t/random-short-touchpad-freezes/30832/28

---
 drivers/pinctrl/pinctrl-amd.c | 7 -------
 1 file changed, 7 deletions(-)
drivers/pinctrl/pinctrl-amd.c +0 -7
index e9b761c2b77a..2d4acf21117c 100644
@@ -468,7 +468,6 @@ static int amd_gpio_irq_set_type(struct irq_data *d, unsigned int type)
 		pin_reg &= ~BIT(LEVEL_TRIG_OFF);
 		pin_reg &= ~(ACTIVE_LEVEL_MASK << ACTIVE_LEVEL_OFF);
 		pin_reg |= ACTIVE_HIGH << ACTIVE_LEVEL_OFF;
-		pin_reg |= DB_TYPE_REMOVE_GLITCH << DB_CNTRL_OFF;
 		irq_set_handler_locked(d, handle_edge_irq);
 		break;
 
@@ -476,7 +475,6 @@ static int amd_gpio_irq_set_type(struct irq_data *d, unsigned int type)
 		pin_reg &= ~BIT(LEVEL_TRIG_OFF);
 		pin_reg &= ~(ACTIVE_LEVEL_MASK << ACTIVE_LEVEL_OFF);
 		pin_reg |= ACTIVE_LOW << ACTIVE_LEVEL_OFF;
-		pin_reg |= DB_TYPE_REMOVE_GLITCH << DB_CNTRL_OFF;
 		irq_set_handler_locked(d, handle_edge_irq);
 		break;
 
@@ -484,7 +482,6 @@ static int amd_gpio_irq_set_type(struct irq_data *d, unsigned int type)
 		pin_reg &= ~BIT(LEVEL_TRIG_OFF);
 		pin_reg &= ~(ACTIVE_LEVEL_MASK << ACTIVE_LEVEL_OFF);
 		pin_reg |= BOTH_EADGE << ACTIVE_LEVEL_OFF;
-		pin_reg |= DB_TYPE_REMOVE_GLITCH << DB_CNTRL_OFF;
 		irq_set_handler_locked(d, handle_edge_irq);
 		break;
 
@@ -492,8 +489,6 @@ static int amd_gpio_irq_set_type(struct irq_data *d, unsigned int type)
 		pin_reg |= LEVEL_TRIGGER << LEVEL_TRIG_OFF;
 		pin_reg &= ~(ACTIVE_LEVEL_MASK << ACTIVE_LEVEL_OFF);
 		pin_reg |= ACTIVE_HIGH << ACTIVE_LEVEL_OFF;
-		pin_reg &= ~(DB_CNTRl_MASK << DB_CNTRL_OFF);
-		pin_reg |= DB_TYPE_PRESERVE_LOW_GLITCH << DB_CNTRL_OFF;
 		irq_set_handler_locked(d, handle_level_irq);
 		break;
 
@@ -501,8 +496,6 @@ static int amd_gpio_irq_set_type(struct irq_data *d, unsigned int type)
 		pin_reg |= LEVEL_TRIGGER << LEVEL_TRIG_OFF;
 		pin_reg &= ~(ACTIVE_LEVEL_MASK << ACTIVE_LEVEL_OFF);
 		pin_reg |= ACTIVE_LOW << ACTIVE_LEVEL_OFF;
-		pin_reg &= ~(DB_CNTRl_MASK << DB_CNTRL_OFF);
-		pin_reg |= DB_TYPE_PRESERVE_HIGH_GLITCH << DB_CNTRL_OFF;
 		irq_set_handler_locked(d, handle_level_irq);
 		break;
 
Signature · 1 line
2.28.0
#10 lore raw

Re: [PATCH v3 4/4] pinctrl: amd: remove debounce filter setting in IRQ type setting

Hi,

Quoted 16 lines · Coiby Xu

Debounce filter setting should be independent from IRQ type setting because according to the ACPI specs, there are separate arguments for specifying debounce timeout and IRQ type in GpioIo() and GpioInt().

This will fix broken touchpads for laptops whose BIOS set the debounce timeout to a relatively large value. For example, the BIOS of Lenovo Legion-5 AMD gaming laptops including 15ARH05 (R7000) and R7000P set the debounce timeout to 124.8ms. This led to the kernel receiving only ~7 HID reports per second from the Synaptics touchpad (MSFT0001:00 06CB:7F28). Existing touchpads like [1][2] are not troubled by this bug because the debounce timeout has been set to 0 by the BIOS before enabling the debounce filter in setting IRQ type.

[1] https://github.com/Syniurge/i2c-amd-mp2/issues/11#issuecomment-721331582
[2] https://forum.manjaro.org/t/random-short-touchpad-freezes/30832/28

Quoted 8 lines

I'm not entirely sure about this patch. This is consistent with how we already stopped touching the debounce timeout setting during init, so that speaks in favor of this change.

Still I'm worried a bit that this might have undesirable side effects.

I guess this should be landed together with Andy's series to apply the debounce setting from the ACPI GPIO resources.

Regards,

Hans

Quoted 51 lines
---
 drivers/pinctrl/pinctrl-amd.c | 7 -------
 1 file changed, 7 deletions(-)
drivers/pinctrl/pinctrl-amd.c +0 -7
index e9b761c2b77a..2d4acf21117c 100644
@@ -468,7 +468,6 @@ static int amd_gpio_irq_set_type(struct irq_data *d, unsigned int type)
 		pin_reg &= ~BIT(LEVEL_TRIG_OFF);
 		pin_reg &= ~(ACTIVE_LEVEL_MASK << ACTIVE_LEVEL_OFF);
 		pin_reg |= ACTIVE_HIGH << ACTIVE_LEVEL_OFF;
-		pin_reg |= DB_TYPE_REMOVE_GLITCH << DB_CNTRL_OFF;
 		irq_set_handler_locked(d, handle_edge_irq);
 		break;
 
@@ -476,7 +475,6 @@ static int amd_gpio_irq_set_type(struct irq_data *d, unsigned int type)
 		pin_reg &= ~BIT(LEVEL_TRIG_OFF);
 		pin_reg &= ~(ACTIVE_LEVEL_MASK << ACTIVE_LEVEL_OFF);
 		pin_reg |= ACTIVE_LOW << ACTIVE_LEVEL_OFF;
-		pin_reg |= DB_TYPE_REMOVE_GLITCH << DB_CNTRL_OFF;
 		irq_set_handler_locked(d, handle_edge_irq);
 		break;
 
@@ -484,7 +482,6 @@ static int amd_gpio_irq_set_type(struct irq_data *d, unsigned int type)
 		pin_reg &= ~BIT(LEVEL_TRIG_OFF);
 		pin_reg &= ~(ACTIVE_LEVEL_MASK << ACTIVE_LEVEL_OFF);
 		pin_reg |= BOTH_EADGE << ACTIVE_LEVEL_OFF;
-		pin_reg |= DB_TYPE_REMOVE_GLITCH << DB_CNTRL_OFF;
 		irq_set_handler_locked(d, handle_edge_irq);
 		break;
 
@@ -492,8 +489,6 @@ static int amd_gpio_irq_set_type(struct irq_data *d, unsigned int type)
 		pin_reg |= LEVEL_TRIGGER << LEVEL_TRIG_OFF;
 		pin_reg &= ~(ACTIVE_LEVEL_MASK << ACTIVE_LEVEL_OFF);
 		pin_reg |= ACTIVE_HIGH << ACTIVE_LEVEL_OFF;
-		pin_reg &= ~(DB_CNTRl_MASK << DB_CNTRL_OFF);
-		pin_reg |= DB_TYPE_PRESERVE_LOW_GLITCH << DB_CNTRL_OFF;
 		irq_set_handler_locked(d, handle_level_irq);
 		break;
 
@@ -501,8 +496,6 @@ static int amd_gpio_irq_set_type(struct irq_data *d, unsigned int type)
 		pin_reg |= LEVEL_TRIGGER << LEVEL_TRIG_OFF;
 		pin_reg &= ~(ACTIVE_LEVEL_MASK << ACTIVE_LEVEL_OFF);
 		pin_reg |= ACTIVE_LOW << ACTIVE_LEVEL_OFF;
-		pin_reg &= ~(DB_CNTRl_MASK << DB_CNTRL_OFF);
-		pin_reg |= DB_TYPE_PRESERVE_HIGH_GLITCH << DB_CNTRL_OFF;
 		irq_set_handler_locked(d, handle_level_irq);
 		break;
 
author
#11 lore raw

Re: [PATCH v3 4/4] pinctrl: amd: remove debounce filter setting in IRQ type setting

Quoted 28 lines · Hans de Goede

Hi,

Quoted 16 lines · Coiby Xu

Debounce filter setting should be independent from IRQ type setting because according to the ACPI specs, there are separate arguments for specifying debounce timeout and IRQ type in GpioIo() and GpioInt().

This will fix broken touchpads for laptops whose BIOS set the debounce timeout to a relatively large value. For example, the BIOS of Lenovo Legion-5 AMD gaming laptops including 15ARH05 (R7000) and R7000P set the debounce timeout to 124.8ms. This led to the kernel receiving only ~7 HID reports per second from the Synaptics touchpad (MSFT0001:00 06CB:7F28). Existing touchpads like [1][2] are not troubled by this bug because the debounce timeout has been set to 0 by the BIOS before enabling the debounce filter in setting IRQ type.

[1] https://github.com/Syniurge/i2c-amd-mp2/issues/11#issuecomment-721331582
[2] https://forum.manjaro.org/t/random-short-touchpad-freezes/30832/28

Quoted 8 lines
Quoted 6 lines

I'm not entirely sure about this patch. This is consistent with how we already stopped touching the debounce timeout setting during init, so that speaks in favor of this change.

Still I'm worried a bit that this might have undesirable side effects.

Now I can only confirm this patch won't affect the mentioned touchpads. I'll see if other distributions like Manjaro are willing to test it through the unstable channel.

Quoted 2 lines

I guess this should be landed together with Andy's series to apply the debounce setting from the ACPI GPIO resources.

Thank you for the reminding! You are right, Andy's patch "gpiolib: acpi: Take into account debounce settings" is needed to fix this kind of touchpad issues. Since that patch hasn't been merged, is there a way to refer to it in the commit message?

Quoted 49 lines

Regards,

Hans

Quoted 41 lines
---
 drivers/pinctrl/pinctrl-amd.c | 7 -------
 1 file changed, 7 deletions(-)
drivers/pinctrl/pinctrl-amd.c +0 -5
index e9b761c2b77a..2d4acf21117c 100644
@@ -468,7 +468,6 @@ static int amd_gpio_irq_set_type(struct irq_data *d, unsigned int type)
 		pin_reg &= ~BIT(LEVEL_TRIG_OFF);
 		pin_reg &= ~(ACTIVE_LEVEL_MASK << ACTIVE_LEVEL_OFF);
 		pin_reg |= ACTIVE_HIGH << ACTIVE_LEVEL_OFF;
-		pin_reg |= DB_TYPE_REMOVE_GLITCH << DB_CNTRL_OFF;
 		irq_set_handler_locked(d, handle_edge_irq);
 		break;
 
@@ -476,7 +475,6 @@ static int amd_gpio_irq_set_type(struct irq_data *d, unsigned int type)
 		pin_reg &= ~BIT(LEVEL_TRIG_OFF);
 		pin_reg &= ~(ACTIVE_LEVEL_MASK << ACTIVE_LEVEL_OFF);
 		pin_reg |= ACTIVE_LOW << ACTIVE_LEVEL_OFF;
-		pin_reg |= DB_TYPE_REMOVE_GLITCH << DB_CNTRL_OFF;
 		irq_set_handler_locked(d, handle_edge_irq);
 		break;
 
@@ -484,7 +482,6 @@ static int amd_gpio_irq_set_type(struct irq_data *d, unsigned int type)
 		pin_reg &= ~BIT(LEVEL_TRIG_OFF);
 		pin_reg &= ~(ACTIVE_LEVEL_MASK << ACTIVE_LEVEL_OFF);
 		pin_reg |= BOTH_EADGE << ACTIVE_LEVEL_OFF;
-		pin_reg |= DB_TYPE_REMOVE_GLITCH << DB_CNTRL_OFF;
 		irq_set_handler_locked(d, handle_edge_irq);
 		break;
 
@@ -492,8 +489,6 @@ static int amd_gpio_irq_set_type(struct irq_data *d, unsigned int type)
 		pin_reg |= LEVEL_TRIGGER << LEVEL_TRIG_OFF;
 		pin_reg &= ~(ACTIVE_LEVEL_MASK << ACTIVE_LEVEL_OFF);
 		pin_reg |= ACTIVE_HIGH << ACTIVE_LEVEL_OFF;
-		pin_reg &= ~(DB_CNTRl_MASK << DB_CNTRL_OFF);
-		pin_reg |= DB_TYPE_PRESERVE_LOW_GLITCH << DB_CNTRL_OFF;
 		irq_set_handler_locked(d, handle_level_irq);
 		break;
 
Quoted 11 lines

drivers/pinctrl/pinctrl-amd.c static int amd_gpio_irq_set_type(struct irq_data *d, unsigned int type) · line 501

Quoted 10 lines
drivers/pinctrl/pinctrl-amd.c +0 -2
@@ -501,8 +496,6 @@ static int amd_gpio_irq_set_type(struct irq_data *d, unsigned int type)
 		pin_reg |= LEVEL_TRIGGER << LEVEL_TRIG_OFF;
 		pin_reg &= ~(ACTIVE_LEVEL_MASK << ACTIVE_LEVEL_OFF);
 		pin_reg |= ACTIVE_LOW << ACTIVE_LEVEL_OFF;
-		pin_reg &= ~(DB_CNTRl_MASK << DB_CNTRL_OFF);
-		pin_reg |= DB_TYPE_PRESERVE_HIGH_GLITCH << DB_CNTRL_OFF;
 		irq_set_handler_locked(d, handle_level_irq);
 		break;
 

--
Best regards,
Coiby

#12 lore raw

Re: [PATCH v3 4/4] pinctrl: amd: remove debounce filter setting in IRQ type setting

Hi,

Quoted 39 lines · Coiby Xu
Quoted 28 lines · Hans de Goede

Hi,

Quoted 16 lines · Coiby Xu

Debounce filter setting should be independent from IRQ type setting because according to the ACPI specs, there are separate arguments for specifying debounce timeout and IRQ type in GpioIo() and GpioInt().

This will fix broken touchpads for laptops whose BIOS set the debounce timeout to a relatively large value. For example, the BIOS of Lenovo Legion-5 AMD gaming laptops including 15ARH05 (R7000) and R7000P set the debounce timeout to 124.8ms. This led to the kernel receiving only ~7 HID reports per second from the Synaptics touchpad (MSFT0001:00 06CB:7F28). Existing touchpads like [1][2] are not troubled by this bug because the debounce timeout has been set to 0 by the BIOS before enabling the debounce filter in setting IRQ type.

[1] https://github.com/Syniurge/i2c-amd-mp2/issues/11#issuecomment-721331582
[2] https://forum.manjaro.org/t/random-short-touchpad-freezes/30832/28

Quoted 8 lines
Quoted 6 lines

I'm not entirely sure about this patch. This is consistent with how we already stopped touching the debounce timeout setting during init, so that speaks in favor of this change.

Still I'm worried a bit that this might have undesirable side effects.

Now I can only confirm this patch won't affect the mentioned touchpads. I'll see if other distributions like Manjaro are willing to test it through the unstable channel.

Quoted 7 lines
Quoted 2 lines

I guess this should be landed together with Andy's series to apply the debounce setting from the ACPI GPIO resources.

Thank you for the reminding! You are right, Andy's patch "gpiolib: acpi: Take into account debounce settings" is needed to fix this kind of touchpad issues. Since that patch hasn't been merged, is there a way to refer to it in the commit message?

You can always refer to it by subject, as you did above.

Regards,

Hans

Quoted 57 lines
Quoted 52 lines
Quoted 51 lines

---
 drivers/pinctrl/pinctrl-amd.c | 7 -------

 1 file changed, 7 deletions(-)
drivers/pinctrl/pinctrl-amd.c +0 -7
index e9b761c2b77a..2d4acf21117c 100644
@@ -468,7 +468,6 @@ static int amd_gpio_irq_set_type(struct irq_data *d, unsigned int type)
          pin_reg &= ~BIT(LEVEL_TRIG_OFF);
          pin_reg &= ~(ACTIVE_LEVEL_MASK << ACTIVE_LEVEL_OFF);
          pin_reg |= ACTIVE_HIGH << ACTIVE_LEVEL_OFF;
-        pin_reg |= DB_TYPE_REMOVE_GLITCH << DB_CNTRL_OFF;
          irq_set_handler_locked(d, handle_edge_irq);
          break;
 
@@ -476,7 +475,6 @@ static int amd_gpio_irq_set_type(struct irq_data *d, unsigned int type)
          pin_reg &= ~BIT(LEVEL_TRIG_OFF);
          pin_reg &= ~(ACTIVE_LEVEL_MASK << ACTIVE_LEVEL_OFF);
          pin_reg |= ACTIVE_LOW << ACTIVE_LEVEL_OFF;
-        pin_reg |= DB_TYPE_REMOVE_GLITCH << DB_CNTRL_OFF;
          irq_set_handler_locked(d, handle_edge_irq);
          break;
 
@@ -484,7 +482,6 @@ static int amd_gpio_irq_set_type(struct irq_data *d, unsigned int type)
          pin_reg &= ~BIT(LEVEL_TRIG_OFF);
          pin_reg &= ~(ACTIVE_LEVEL_MASK << ACTIVE_LEVEL_OFF);
          pin_reg |= BOTH_EADGE << ACTIVE_LEVEL_OFF;
-        pin_reg |= DB_TYPE_REMOVE_GLITCH << DB_CNTRL_OFF;
          irq_set_handler_locked(d, handle_edge_irq);
          break;
 
@@ -492,8 +489,6 @@ static int amd_gpio_irq_set_type(struct irq_data *d, unsigned int type)
          pin_reg |= LEVEL_TRIGGER << LEVEL_TRIG_OFF;
          pin_reg &= ~(ACTIVE_LEVEL_MASK << ACTIVE_LEVEL_OFF);
          pin_reg |= ACTIVE_HIGH << ACTIVE_LEVEL_OFF;
-        pin_reg &= ~(DB_CNTRl_MASK << DB_CNTRL_OFF);
-        pin_reg |= DB_TYPE_PRESERVE_LOW_GLITCH << DB_CNTRL_OFF;
          irq_set_handler_locked(d, handle_level_irq);
          break;
 
@@ -501,8 +496,6 @@ static int amd_gpio_irq_set_type(struct irq_data *d, unsigned int type)
          pin_reg |= LEVEL_TRIGGER << LEVEL_TRIG_OFF;
          pin_reg &= ~(ACTIVE_LEVEL_MASK << ACTIVE_LEVEL_OFF);
          pin_reg |= ACTIVE_LOW << ACTIVE_LEVEL_OFF;
-        pin_reg &= ~(DB_CNTRl_MASK << DB_CNTRL_OFF);
-        pin_reg |= DB_TYPE_PRESERVE_HIGH_GLITCH << DB_CNTRL_OFF;
          irq_set_handler_locked(d, handle_level_irq);
          break;
 
Signature · 2 lines
Best regards,
Coiby
Acked-by on v3 4/4
#13 lore raw

Re: [PATCH v3 4/4] pinctrl: amd: remove debounce filter setting in IRQ type setting

In reply to Coiby Xu
Quoted 16 lines · Coiby Xu

Debounce filter setting should be independent from IRQ type setting because according to the ACPI specs, there are separate arguments for specifying debounce timeout and IRQ type in GpioIo() and GpioInt().

This will fix broken touchpads for laptops whose BIOS set the debounce timeout to a relatively large value. For example, the BIOS of Lenovo Legion-5 AMD gaming laptops including 15ARH05 (R7000) and R7000P set the debounce timeout to 124.8ms. This led to the kernel receiving only ~7 HID reports per second from the Synaptics touchpad (MSFT0001:00 06CB:7F28). Existing touchpads like [1][2] are not troubled by this bug because the debounce timeout has been set to 0 by the BIOS before enabling the debounce filter in setting IRQ type.

[1] https://github.com/Syniurge/i2c-amd-mp2/issues/11#issuecomment-721331582
[2] https://forum.manjaro.org/t/random-short-touchpad-freezes/30832/28

Quoted 8 lines

As I have applied patches 1-3 we only have this one to land.

If Andy or someone else needs to take it through the ACPI tree you can add my:

If I should apply it or if Andy sends me a pull request, just ping me and tell me what to do :)

Yours,
Linus Walleij

#14 lore raw

Re: [PATCH v3 4/4] pinctrl: amd: remove debounce filter setting in IRQ type setting

Quoted 1 line · Linus Walleij

On Fri, Nov 6, 2020 at 12:19 AM Coiby Xu <[email protected]> wrote:

...

Quoted 6 lines

If Andy or someone else needs to take it through the ACPI tree you can add my:

If I should apply it or if Andy sends me a pull request, just ping me and tell me what to do :)

I can take it, but I would need few things:
- Hans' blessing of my series
- tested on top of my series (hence tested-by tag for my series)
- I would like to perform more tests myself

Signature · 2 lines
With Best Regards,
Andy Shevchenko