Skip to contentSkip to conversation

irqchip/stm32: Keep pinctrl block clock enabled when LEVEL IRQ requested

PATCH Superseded Marek Vasut 4 participants 7 messages Apr 21, 2022 – Apr 22, 2022

Subsystems (4)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 irqchip/stm32: Keep pinctrl block clock enabled when LEVEL IRQ requested

The current EOI handler for LEVEL triggered interrupts calls clk_enable(), register IO, clk_disable(). The clock manipulation requires locking which happens with IRQs disabled in clk_enable_lock(). Instead of turning the clock on and off all the time, enable the clock in case LEVEL interrupt is requested and keep the clock enabled until all LEVEL interrupts are freed. The LEVEL interrupts are an exception on this platform and seldom used, so this does not affect the common case.

This simplifies the LEVEL interrupt handling considerably and also fixes the following splat found when using preempt-rt:

 ------------[ cut here ]------------
 WARNING: CPU: 0 PID: 0 at kernel/locking/rtmutex.c:2040 __rt_mutex_trylock+0x37/0x62
 Modules linked in:
 CPU: 0 PID: 0 Comm: swapper/0 Not tainted 5.10.109-rt65-stable-standard-00068-g6a5afc4b1217 #85
 Hardware name: STM32 (Device Tree Support)
 [<c010a45d>] (unwind_backtrace) from [<c010766f>] (show_stack+0xb/0xc)
 [<c010766f>] (show_stack) from [<c06353ab>] (dump_stack+0x6f/0x84)
 [<c06353ab>] (dump_stack) from [<c01145e3>] (__warn+0x7f/0xa4)
 [<c01145e3>] (__warn) from [<c063386f>] (warn_slowpath_fmt+0x3b/0x74)
 [<c063386f>] (warn_slowpath_fmt) from [<c063b43d>] (__rt_mutex_trylock+0x37/0x62)
 [<c063b43d>] (__rt_mutex_trylock) from [<c063c053>] (rt_spin_trylock+0x7/0x16)
 [<c063c053>] (rt_spin_trylock) from [<c036a2f3>] (clk_enable_lock+0xb/0x80)
 [<c036a2f3>] (clk_enable_lock) from [<c036ba69>] (clk_core_enable_lock+0x9/0x18)
 [<c036ba69>] (clk_core_enable_lock) from [<c034e9f3>] (stm32_gpio_get+0x11/0x24)
 [<c034e9f3>] (stm32_gpio_get) from [<c034ef43>] (stm32_gpio_irq_trigger+0x1f/0x48)
 [<c034ef43>] (stm32_gpio_irq_trigger) from [<c014aa53>] (handle_fasteoi_irq+0x71/0xa8)
 [<c014aa53>] (handle_fasteoi_irq) from [<c0147111>] (generic_handle_irq+0x19/0x22)
 [<c0147111>] (generic_handle_irq) from [<c014752d>] (__handle_domain_irq+0x55/0x64)
 [<c014752d>] (__handle_domain_irq) from [<c0346f13>] (gic_handle_irq+0x53/0x64)
 [<c0346f13>] (gic_handle_irq) from [<c0100ba5>] (__irq_svc+0x65/0xc0)
 Exception stack(0xc0e01f18 to 0xc0e01f60)
 1f00:                                                       0000300c 00000000
 1f20: 0000300c c010ff01 00000000 00000000 c0e00000 c0e07714 00000001 c0e01f78
 1f40: c0e07758 00000000 ef7cd0ff c0e01f68 c010554b c0105542 40000033 ffffffff
 [<c0100ba5>] (__irq_svc) from [<c0105542>] (arch_cpu_idle+0xc/0x1e)
 [<c0105542>] (arch_cpu_idle) from [<c063be95>] (default_idle_call+0x21/0x3c)
 [<c063be95>] (default_idle_call) from [<c01324f7>] (do_idle+0xe3/0x1e4)
 [<c01324f7>] (do_idle) from [<c01327b3>] (cpu_startup_entry+0x13/0x14)
 [<c01327b3>] (cpu_startup_entry) from [<c0a00c13>] (start_kernel+0x397/0x3d4)
 [<c0a00c13>] (start_kernel) from [<00000000>] (0x0)
 ---[ end trace 0000000000000002 ]---

Power consumption measured on STM32MP157C DHCOM SoM is not increased or is below noise threshold.

Fixes 47beed513a85b ("pinctrl: stm32: Add level interrupt support to gpio irq chip")
Cc Alexandre Torgue <[email protected]>
Cc Fabien Dessenne <[email protected]>
Cc Linus Walleij <[email protected]>
Cc Marc Zyngier <[email protected]>

To: [email protected]

---
 drivers/pinctrl/stm32/pinctrl-stm32.c | 19 +++++++++++++++++--
 1 file changed, 17 insertions(+), 2 deletions(-)
drivers/pinctrl/stm32/pinctrl-stm32.c +17 -2
index 242d1c37c6e4..7aecd0efde07 100644
@@ -226,6 +226,13 @@ static void stm32_gpio_free(struct gpio_chip *chip, unsigned offset)
 	pinctrl_gpio_free(chip->base + offset);
 }
 
+static int stm32_gpio_get_noclk(struct gpio_chip *chip, unsigned int offset)
+{
+	struct stm32_gpio_bank *bank = gpiochip_get_data(chip);
+
+	return !!(readl_relaxed(bank->base + STM32_GPIO_IDR) & BIT(offset));
+}
+
 static int stm32_gpio_get(struct gpio_chip *chip, unsigned offset)
 {
 	struct stm32_gpio_bank *bank = gpiochip_get_data(chip);
@@ -233,7 +240,7 @@ static int stm32_gpio_get(struct gpio_chip *chip, unsigned offset)
 
 	clk_enable(bank->clk);
 
-	ret = !!(readl_relaxed(bank->base + STM32_GPIO_IDR) & BIT(offset));
+	ret = stm32_gpio_get_noclk(chip, offset);
 
 	clk_disable(bank->clk);
 
@@ -317,7 +324,7 @@ static void stm32_gpio_irq_trigger(struct irq_data *d)
 		return;
 
 	/* If level interrupt type then retrig */
-	level = stm32_gpio_get(&bank->gpio_chip, d->hwirq);
+	level = stm32_gpio_get_noclk(&bank->gpio_chip, d->hwirq);
 	if ((level == 0 && bank->irq_type[d->hwirq] == IRQ_TYPE_LEVEL_LOW) ||
 	    (level == 1 && bank->irq_type[d->hwirq] == IRQ_TYPE_LEVEL_HIGH))
 		irq_chip_retrigger_hierarchy(d);
@@ -359,6 +366,7 @@ static int stm32_gpio_irq_request_resources(struct irq_data *irq_data)
 {
 	struct stm32_gpio_bank *bank = irq_data->domain->host_data;
 	struct stm32_pinctrl *pctl = dev_get_drvdata(bank->gpio_chip.parent);
+	unsigned long flags;
 	int ret;
 
 	ret = stm32_gpio_direction_input(&bank->gpio_chip, irq_data->hwirq);
@@ -372,6 +380,10 @@ static int stm32_gpio_irq_request_resources(struct irq_data *irq_data)
 		return ret;
 	}
 
+	flags = irqd_get_trigger_type(irq_data);
+	if (flags & IRQ_TYPE_LEVEL_MASK)
+		clk_enable(bank->clk);
+
 	return 0;
 }
 
@@ -379,6 +391,9 @@ static void stm32_gpio_irq_release_resources(struct irq_data *irq_data)
 {
 	struct stm32_gpio_bank *bank = irq_data->domain->host_data;
 
+	if (bank->irq_type[irq_data->hwirq] & IRQ_TYPE_LEVEL_MASK)
+		clk_disable(bank->clk);
+
 	gpiochip_unlock_as_irq(&bank->gpio_chip, irq_data->hwirq);
 }
 
Signature · 1 line
2.35.1
#1 lore raw

Hi Marek,

I agree there is something wrong with the clock management in IRQ context here and your patch goes in the right way. There are also some other problems regarding performance (enabling / disabling clock each time we want to change the IO value, ...). For these both issues I have a patch, which basically keeps the GPIO clocks enabled from probe.
I did not have time to submit it, but, considering your concerns, I will do it in the coming days.
For the time being I suggest that we do not apply your patch.

BR
Fabien

Quoted 123 lines · Marek Vasut

The current EOI handler for LEVEL triggered interrupts calls clk_enable(), register IO, clk_disable(). The clock manipulation requires locking which happens with IRQs disabled in clk_enable_lock(). Instead of turning the clock on and off all the time, enable the clock in case LEVEL interrupt is requested and keep the clock enabled until all LEVEL interrupts are freed. The LEVEL interrupts are an exception on this platform and seldom used, so this does not affect the common case.

This simplifies the LEVEL interrupt handling considerably and also fixes the following splat found when using preempt-rt:

  ------------[ cut here ]------------
  WARNING: CPU: 0 PID: 0 at kernel/locking/rtmutex.c:2040 __rt_mutex_trylock+0x37/0x62
  Modules linked in:
  CPU: 0 PID: 0 Comm: swapper/0 Not tainted 5.10.109-rt65-stable-standard-00068-g6a5afc4b1217 #85
  Hardware name: STM32 (Device Tree Support)
  [<c010a45d>] (unwind_backtrace) from [<c010766f>] (show_stack+0xb/0xc)
  [<c010766f>] (show_stack) from [<c06353ab>] (dump_stack+0x6f/0x84)
  [<c06353ab>] (dump_stack) from [<c01145e3>] (__warn+0x7f/0xa4)
  [<c01145e3>] (__warn) from [<c063386f>] (warn_slowpath_fmt+0x3b/0x74)
  [<c063386f>] (warn_slowpath_fmt) from [<c063b43d>] (__rt_mutex_trylock+0x37/0x62)
  [<c063b43d>] (__rt_mutex_trylock) from [<c063c053>] (rt_spin_trylock+0x7/0x16)
  [<c063c053>] (rt_spin_trylock) from [<c036a2f3>] (clk_enable_lock+0xb/0x80)
  [<c036a2f3>] (clk_enable_lock) from [<c036ba69>] (clk_core_enable_lock+0x9/0x18)
  [<c036ba69>] (clk_core_enable_lock) from [<c034e9f3>] (stm32_gpio_get+0x11/0x24)
  [<c034e9f3>] (stm32_gpio_get) from [<c034ef43>] (stm32_gpio_irq_trigger+0x1f/0x48)
  [<c034ef43>] (stm32_gpio_irq_trigger) from [<c014aa53>] (handle_fasteoi_irq+0x71/0xa8)
  [<c014aa53>] (handle_fasteoi_irq) from [<c0147111>] (generic_handle_irq+0x19/0x22)
  [<c0147111>] (generic_handle_irq) from [<c014752d>] (__handle_domain_irq+0x55/0x64)
  [<c014752d>] (__handle_domain_irq) from [<c0346f13>] (gic_handle_irq+0x53/0x64)
  [<c0346f13>] (gic_handle_irq) from [<c0100ba5>] (__irq_svc+0x65/0xc0)
  Exception stack(0xc0e01f18 to 0xc0e01f60)
  1f00:                                                       0000300c 00000000
  1f20: 0000300c c010ff01 00000000 00000000 c0e00000 c0e07714 00000001 c0e01f78
  1f40: c0e07758 00000000 ef7cd0ff c0e01f68 c010554b c0105542 40000033 ffffffff
  [<c0100ba5>] (__irq_svc) from [<c0105542>] (arch_cpu_idle+0xc/0x1e)
  [<c0105542>] (arch_cpu_idle) from [<c063be95>] (default_idle_call+0x21/0x3c)
  [<c063be95>] (default_idle_call) from [<c01324f7>] (do_idle+0xe3/0x1e4)
  [<c01324f7>] (do_idle) from [<c01327b3>] (cpu_startup_entry+0x13/0x14)
  [<c01327b3>] (cpu_startup_entry) from [<c0a00c13>] (start_kernel+0x397/0x3d4)
  [<c0a00c13>] (start_kernel) from [<00000000>] (0x0)
  ---[ end trace 0000000000000002 ]---

Power consumption measured on STM32MP157C DHCOM SoM is not increased or is below noise threshold.

Fixes 47beed513a85b ("pinctrl: stm32: Add level interrupt support to gpio irq chip")
Cc Alexandre Torgue <[email protected]>
Cc Fabien Dessenne <[email protected]>
Cc Linus Walleij <[email protected]>
Cc Marc Zyngier <[email protected]>

To: [email protected]

---
  drivers/pinctrl/stm32/pinctrl-stm32.c | 19 +++++++++++++++++--
  1 file changed, 17 insertions(+), 2 deletions(-)
drivers/pinctrl/stm32/pinctrl-stm32.c +17 -2
index 242d1c37c6e4..7aecd0efde07 100644
@@ -226,6 +226,13 @@ static void stm32_gpio_free(struct gpio_chip *chip, unsigned offset)
  	pinctrl_gpio_free(chip->base + offset);
  }
 
+static int stm32_gpio_get_noclk(struct gpio_chip *chip, unsigned int offset)
+{
+	struct stm32_gpio_bank *bank = gpiochip_get_data(chip);
+
+	return !!(readl_relaxed(bank->base + STM32_GPIO_IDR) & BIT(offset));
+}
+
  static int stm32_gpio_get(struct gpio_chip *chip, unsigned offset)
  {
  	struct stm32_gpio_bank *bank = gpiochip_get_data(chip);
@@ -233,7 +240,7 @@ static int stm32_gpio_get(struct gpio_chip *chip, unsigned offset)
 
  	clk_enable(bank->clk);
 
-	ret = !!(readl_relaxed(bank->base + STM32_GPIO_IDR) & BIT(offset));
+	ret = stm32_gpio_get_noclk(chip, offset);
 
  	clk_disable(bank->clk);
 
@@ -317,7 +324,7 @@ static void stm32_gpio_irq_trigger(struct irq_data *d)
  		return;
 
  	/* If level interrupt type then retrig */
-	level = stm32_gpio_get(&bank->gpio_chip, d->hwirq);
+	level = stm32_gpio_get_noclk(&bank->gpio_chip, d->hwirq);
  	if ((level == 0 && bank->irq_type[d->hwirq] == IRQ_TYPE_LEVEL_LOW) ||
  	    (level == 1 && bank->irq_type[d->hwirq] == IRQ_TYPE_LEVEL_HIGH))
  		irq_chip_retrigger_hierarchy(d);
@@ -359,6 +366,7 @@ static int stm32_gpio_irq_request_resources(struct irq_data *irq_data)
  {
  	struct stm32_gpio_bank *bank = irq_data->domain->host_data;
  	struct stm32_pinctrl *pctl = dev_get_drvdata(bank->gpio_chip.parent);
+	unsigned long flags;
  	int ret;
 
  	ret = stm32_gpio_direction_input(&bank->gpio_chip, irq_data->hwirq);
@@ -372,6 +380,10 @@ static int stm32_gpio_irq_request_resources(struct irq_data *irq_data)
  		return ret;
  	}
 
+	flags = irqd_get_trigger_type(irq_data);
+	if (flags & IRQ_TYPE_LEVEL_MASK)
+		clk_enable(bank->clk);
+
  	return 0;
  }
 
@@ -379,6 +391,9 @@ static void stm32_gpio_irq_release_resources(struct irq_data *irq_data)
  {
  	struct stm32_gpio_bank *bank = irq_data->domain->host_data;
 
+	if (bank->irq_type[irq_data->hwirq] & IRQ_TYPE_LEVEL_MASK)
+		clk_disable(bank->clk);
+
  	gpiochip_unlock_as_irq(&bank->gpio_chip, irq_data->hwirq);
  }
 
#2 lore raw
Quoted 1 line · Fabien DESSENNE

Hi Marek,

Hi,

Quoted 2 lines

I agree there is something wrong with the clock management in IRQ context here and your patch goes in the right way. There are also some other problems regarding performance (enabling / disabling clock each time we want to change the IO value, ...).

These are already fixed by separate patch:
[PATCH] irqchip/stm32: Do not call stm32_gpio_get() for edge triggered IRQs in EOI

Quoted 3 lines

For these both issues I have a patch, which basically keeps the GPIO clocks enabled from probe. I did not have time to submit it, but, considering your concerns, I will do it in the coming days. For the time being I suggest that we do not apply your patch.

Is there anything specifically wrong with this patch ?

I think you should be easily able to apply your patch which keeps the clock enabled all the time on top of this one. The added bonus is that the commit message of this patch - which explains what the problem is with the LEVEL IRQs - would be retained, and so would bisectability in case your patch causes some sort of problem.

Quoted 123 lines

BR
Fabien

Quoted 119 lines · Marek Vasut

The current EOI handler for LEVEL triggered interrupts calls clk_enable(), register IO, clk_disable(). The clock manipulation requires locking which happens with IRQs disabled in clk_enable_lock(). Instead of turning the clock on and off all the time, enable the clock in case LEVEL interrupt is requested and keep the clock enabled until all LEVEL interrupts are freed. The LEVEL interrupts are an exception on this platform and seldom used, so this does not affect the common case.

This simplifies the LEVEL interrupt handling considerably and also fixes the following splat found when using preempt-rt:

  ------------[ cut here ]------------
  WARNING: CPU: 0 PID: 0 at kernel/locking/rtmutex.c:2040 __rt_mutex_trylock+0x37/0x62
  Modules linked in:
  CPU: 0 PID: 0 Comm: swapper/0 Not tainted 5.10.109-rt65-stable-standard-00068-g6a5afc4b1217 #85
  Hardware name: STM32 (Device Tree Support)
  [<c010a45d>] (unwind_backtrace) from [<c010766f>] (show_stack+0xb/0xc)
  [<c010766f>] (show_stack) from [<c06353ab>] (dump_stack+0x6f/0x84)
  [<c06353ab>] (dump_stack) from [<c01145e3>] (__warn+0x7f/0xa4)
  [<c01145e3>] (__warn) from [<c063386f>] (warn_slowpath_fmt+0x3b/0x74)
  [<c063386f>] (warn_slowpath_fmt) from [<c063b43d>] (__rt_mutex_trylock+0x37/0x62)
  [<c063b43d>] (__rt_mutex_trylock) from [<c063c053>] (rt_spin_trylock+0x7/0x16)
  [<c063c053>] (rt_spin_trylock) from [<c036a2f3>] (clk_enable_lock+0xb/0x80)
  [<c036a2f3>] (clk_enable_lock) from [<c036ba69>] (clk_core_enable_lock+0x9/0x18)
  [<c036ba69>] (clk_core_enable_lock) from [<c034e9f3>] (stm32_gpio_get+0x11/0x24)
  [<c034e9f3>] (stm32_gpio_get) from [<c034ef43>] (stm32_gpio_irq_trigger+0x1f/0x48)
  [<c034ef43>] (stm32_gpio_irq_trigger) from [<c014aa53>] (handle_fasteoi_irq+0x71/0xa8)
  [<c014aa53>] (handle_fasteoi_irq) from [<c0147111>] (generic_handle_irq+0x19/0x22)
  [<c0147111>] (generic_handle_irq) from [<c014752d>] (__handle_domain_irq+0x55/0x64)
  [<c014752d>] (__handle_domain_irq) from [<c0346f13>] (gic_handle_irq+0x53/0x64)
  [<c0346f13>] (gic_handle_irq) from [<c0100ba5>] (__irq_svc+0x65/0xc0)
  Exception stack(0xc0e01f18 to 0xc0e01f60)

  1f00:                                                       0000300c 00000000 1f20: 0000300c c010ff01 00000000 00000000 c0e00000 c0e07714 00000001 c0e01f78 1f40: c0e07758 00000000 ef7cd0ff c0e01f68 c010554b c0105542 40000033 ffffffff [<c0100ba5>] (__irq_svc) from [<c0105542>] (arch_cpu_idle+0xc/0x1e) [<c0105542>] (arch_cpu_idle) from [<c063be95>] (default_idle_call+0x21/0x3c) [<c063be95>] (default_idle_call) from [<c01324f7>] (do_idle+0xe3/0x1e4) [<c01324f7>] (do_idle) from [<c01327b3>] (cpu_startup_entry+0x13/0x14) [<c01327b3>] (cpu_startup_entry) from [<c0a00c13>] (start_kernel+0x397/0x3d4) [<c0a00c13>] (start_kernel) from [<00000000>] (0x0)

  ---[ end trace 0000000000000002 ]---

Power consumption measured on STM32MP157C DHCOM SoM is not increased or is below noise threshold.

Fixes 47beed513a85b ("pinctrl: stm32: Add level interrupt support to gpio irq chip")
Cc Alexandre Torgue <[email protected]>
Cc Fabien Dessenne <[email protected]>
Cc Linus Walleij <[email protected]>
Cc Marc Zyngier <[email protected]>

To: [email protected]
---
  drivers/pinctrl/stm32/pinctrl-stm32.c | 19 +++++++++++++++++--

  1 file changed, 17 insertions(+), 2 deletions(-)
drivers/pinctrl/stm32/pinctrl-stm32.c +7 -0
index 242d1c37c6e4..7aecd0efde07 100644
@@ -226,6 +226,13 @@ static void stm32_gpio_free(struct gpio_chip 
 *chip, unsigned offset)
       pinctrl_gpio_free(chip->base + offset);
   }
+static int stm32_gpio_get_noclk(struct gpio_chip *chip, unsigned int 
 offset)
+{
+    struct stm32_gpio_bank *bank = gpiochip_get_data(chip);
+
+    return !!(readl_relaxed(bank->base + STM32_GPIO_IDR) & BIT(offset));
+}
+
   static int stm32_gpio_get(struct gpio_chip *chip, unsigned offset)
   {

      struct stm32_gpio_bank *bank = gpiochip_get_data(chip);

drivers/pinctrl/stm32/pinctrl-stm32.c +2 -2
@@ -233,7 +240,7 @@ static int stm32_gpio_get(struct gpio_chip *chip, 
 unsigned offset)
       clk_enable(bank->clk);
-    ret = !!(readl_relaxed(bank->base + STM32_GPIO_IDR) & BIT(offset));
+    ret = stm32_gpio_get_noclk(chip, offset);
       clk_disable(bank->clk);
@@ -317,7 +324,7 @@ static void stm32_gpio_irq_trigger(struct irq_data 
 *d)
           return;
       /* If level interrupt type then retrig */
-    level = stm32_gpio_get(&bank->gpio_chip, d->hwirq);
+    level = stm32_gpio_get_noclk(&bank->gpio_chip, d->hwirq);

      if ((level == 0 && bank->irq_type[d->hwirq] == IRQ_TYPE_LEVEL_LOW) || (level == 1 && bank->irq_type[d->hwirq] == IRQ_TYPE_LEVEL_HIGH)) irq_chip_retrigger_hierarchy(d);

drivers/pinctrl/stm32/pinctrl-stm32.c +8 -0
@@ -359,6 +366,7 @@ static int stm32_gpio_irq_request_resources(struct 
 irq_data *irq_data)
   {
       struct stm32_gpio_bank *bank = irq_data->domain->host_data;
       struct stm32_pinctrl *pctl = dev_get_drvdata(bank->gpio_chip.parent);
+    unsigned long flags;
       int ret;
       ret = stm32_gpio_direction_input(&bank->gpio_chip, irq_data->hwirq);
@@ -372,6 +380,10 @@ static int 
 stm32_gpio_irq_request_resources(struct irq_data *irq_data)
           return ret;
       }
+    flags = irqd_get_trigger_type(irq_data);
+    if (flags & IRQ_TYPE_LEVEL_MASK)
+        clk_enable(bank->clk);
+
       return 0;
   }
@@ -379,6 +391,9 @@ static void 
 stm32_gpio_irq_release_resources(struct irq_data *irq_data)
   {
       struct stm32_gpio_bank *bank = irq_data->domain->host_data;
+    if (bank->irq_type[irq_data->hwirq] & IRQ_TYPE_LEVEL_MASK)
+        clk_disable(bank->clk);
+
       gpiochip_unlock_as_irq(&bank->gpio_chip, irq_data->hwirq);
   }
#3 lore raw
In reply to Fabien DESSENNE
Quoted 12 lines · Fabien DESSENNE

Hi Marek,

I agree there is something wrong with the clock management in IRQ context here and your patch goes in the right way.
There are also some other problems regarding performance (enabling / disabling clock each time we want to change the IO value, ...). For these both issues I have a patch, which basically keeps the GPIO clocks enabled from probe.
I did not have time to submit it, but, considering your concerns, I will do it in the coming days.
For the time being I suggest that we do not apply your patch.

Why? This fixes a glaring issue, and there are no alternatives at the moment. So if there is something to improve on, please base your patch on top of Marek's.

Thanks,

	M.
Signature · 1 line
Without deviation from the norm, progress is not possible.
#4 lore raw

Hi Mark and Marek,

My intention was to have a single commit, instead of two commits with the second one making a part of the previous one obsolete.

Nevertheless I understand your points which are valid, so I will review Marek's patch first before submitting another patch.

BR
Fabien

Quoted 23 lines · Marc Zyngier
Quoted 12 lines · Fabien DESSENNE

Hi Marek,

I agree there is something wrong with the clock management in IRQ context here and your patch goes in the right way.
There are also some other problems regarding performance (enabling / disabling clock each time we want to change the IO value, ...). For these both issues I have a patch, which basically keeps the GPIO clocks enabled from probe.
I did not have time to submit it, but, considering your concerns, I will do it in the coming days.
For the time being I suggest that we do not apply your patch.

Why? This fixes a glaring issue, and there are no alternatives at the moment. So if there is something to improve on, please base your patch on top of Marek's.

Thanks,

	M.
Reviewed-by on v1 1/1
#5 lore raw
In reply to Marek Vasut

Hi Marek

Thank you for the patch.

Like Linus did for your other patch, I would prefer the subject prefix to be "pinctrl: stm32:"

Apart from that I added my RB below

BR
Fabien

Quoted 42 lines · Marek Vasut

The current EOI handler for LEVEL triggered interrupts calls clk_enable(), register IO, clk_disable(). The clock manipulation requires locking which happens with IRQs disabled in clk_enable_lock(). Instead of turning the clock on and off all the time, enable the clock in case LEVEL interrupt is requested and keep the clock enabled until all LEVEL interrupts are freed. The LEVEL interrupts are an exception on this platform and seldom used, so this does not affect the common case.

This simplifies the LEVEL interrupt handling considerably and also fixes the following splat found when using preempt-rt:

  ------------[ cut here ]------------
  WARNING: CPU: 0 PID: 0 at kernel/locking/rtmutex.c:2040 __rt_mutex_trylock+0x37/0x62
  Modules linked in:
  CPU: 0 PID: 0 Comm: swapper/0 Not tainted 5.10.109-rt65-stable-standard-00068-g6a5afc4b1217 #85
  Hardware name: STM32 (Device Tree Support)
  [<c010a45d>] (unwind_backtrace) from [<c010766f>] (show_stack+0xb/0xc)
  [<c010766f>] (show_stack) from [<c06353ab>] (dump_stack+0x6f/0x84)
  [<c06353ab>] (dump_stack) from [<c01145e3>] (__warn+0x7f/0xa4)
  [<c01145e3>] (__warn) from [<c063386f>] (warn_slowpath_fmt+0x3b/0x74)
  [<c063386f>] (warn_slowpath_fmt) from [<c063b43d>] (__rt_mutex_trylock+0x37/0x62)
  [<c063b43d>] (__rt_mutex_trylock) from [<c063c053>] (rt_spin_trylock+0x7/0x16)
  [<c063c053>] (rt_spin_trylock) from [<c036a2f3>] (clk_enable_lock+0xb/0x80)
  [<c036a2f3>] (clk_enable_lock) from [<c036ba69>] (clk_core_enable_lock+0x9/0x18)
  [<c036ba69>] (clk_core_enable_lock) from [<c034e9f3>] (stm32_gpio_get+0x11/0x24)
  [<c034e9f3>] (stm32_gpio_get) from [<c034ef43>] (stm32_gpio_irq_trigger+0x1f/0x48)
  [<c034ef43>] (stm32_gpio_irq_trigger) from [<c014aa53>] (handle_fasteoi_irq+0x71/0xa8)
  [<c014aa53>] (handle_fasteoi_irq) from [<c0147111>] (generic_handle_irq+0x19/0x22)
  [<c0147111>] (generic_handle_irq) from [<c014752d>] (__handle_domain_irq+0x55/0x64)
  [<c014752d>] (__handle_domain_irq) from [<c0346f13>] (gic_handle_irq+0x53/0x64)
  [<c0346f13>] (gic_handle_irq) from [<c0100ba5>] (__irq_svc+0x65/0xc0)
  Exception stack(0xc0e01f18 to 0xc0e01f60)
  1f00:                                                       0000300c 00000000
  1f20: 0000300c c010ff01 00000000 00000000 c0e00000 c0e07714 00000001 c0e01f78
  1f40: c0e07758 00000000 ef7cd0ff c0e01f68 c010554b c0105542 40000033 ffffffff
  [<c0100ba5>] (__irq_svc) from [<c0105542>] (arch_cpu_idle+0xc/0x1e)
  [<c0105542>] (arch_cpu_idle) from [<c063be95>] (default_idle_call+0x21/0x3c)
  [<c063be95>] (default_idle_call) from [<c01324f7>] (do_idle+0xe3/0x1e4)
  [<c01324f7>] (do_idle) from [<c01327b3>] (cpu_startup_entry+0x13/0x14)
  [<c01327b3>] (cpu_startup_entry) from [<c0a00c13>] (start_kernel+0x397/0x3d4)
  [<c0a00c13>] (start_kernel) from [<00000000>] (0x0)
  ---[ end trace 0000000000000002 ]---
Quoted 5 lines

Power consumption measured on STM32MP157C DHCOM SoM is not increased or is below noise threshold.

Fixes 47beed513a85b ("pinctrl: stm32: Add level interrupt support to gpio irq chip")
Quoted 76 lines

Cc: Alexandre Torgue <[email protected]>
Cc: Fabien Dessenne <[email protected]>
Cc: Linus Walleij <[email protected]>
Cc: Marc Zyngier <[email protected]>
Cc: [email protected]
Cc: [email protected]
To: [email protected]

---
  drivers/pinctrl/stm32/pinctrl-stm32.c | 19 +++++++++++++++++--
  1 file changed, 17 insertions(+), 2 deletions(-)
drivers/pinctrl/stm32/pinctrl-stm32.c +17 -2
index 242d1c37c6e4..7aecd0efde07 100644
@@ -226,6 +226,13 @@ static void stm32_gpio_free(struct gpio_chip *chip, unsigned offset)
  	pinctrl_gpio_free(chip->base + offset);
  }
 
+static int stm32_gpio_get_noclk(struct gpio_chip *chip, unsigned int offset)
+{
+	struct stm32_gpio_bank *bank = gpiochip_get_data(chip);
+
+	return !!(readl_relaxed(bank->base + STM32_GPIO_IDR) & BIT(offset));
+}
+
  static int stm32_gpio_get(struct gpio_chip *chip, unsigned offset)
  {
  	struct stm32_gpio_bank *bank = gpiochip_get_data(chip);
@@ -233,7 +240,7 @@ static int stm32_gpio_get(struct gpio_chip *chip, unsigned offset)
 
  	clk_enable(bank->clk);
 
-	ret = !!(readl_relaxed(bank->base + STM32_GPIO_IDR) & BIT(offset));
+	ret = stm32_gpio_get_noclk(chip, offset);
 
  	clk_disable(bank->clk);
 
@@ -317,7 +324,7 @@ static void stm32_gpio_irq_trigger(struct irq_data *d)
  		return;
 
  	/* If level interrupt type then retrig */
-	level = stm32_gpio_get(&bank->gpio_chip, d->hwirq);
+	level = stm32_gpio_get_noclk(&bank->gpio_chip, d->hwirq);
  	if ((level == 0 && bank->irq_type[d->hwirq] == IRQ_TYPE_LEVEL_LOW) ||
  	    (level == 1 && bank->irq_type[d->hwirq] == IRQ_TYPE_LEVEL_HIGH))
  		irq_chip_retrigger_hierarchy(d);
@@ -359,6 +366,7 @@ static int stm32_gpio_irq_request_resources(struct irq_data *irq_data)
  {
  	struct stm32_gpio_bank *bank = irq_data->domain->host_data;
  	struct stm32_pinctrl *pctl = dev_get_drvdata(bank->gpio_chip.parent);
+	unsigned long flags;
  	int ret;
 
  	ret = stm32_gpio_direction_input(&bank->gpio_chip, irq_data->hwirq);
@@ -372,6 +380,10 @@ static int stm32_gpio_irq_request_resources(struct irq_data *irq_data)
  		return ret;
  	}
 
+	flags = irqd_get_trigger_type(irq_data);
+	if (flags & IRQ_TYPE_LEVEL_MASK)
+		clk_enable(bank->clk);
+
  	return 0;
  }
 
@@ -379,6 +391,9 @@ static void stm32_gpio_irq_release_resources(struct irq_data *irq_data)
  {
  	struct stm32_gpio_bank *bank = irq_data->domain->host_data;
 
+	if (bank->irq_type[irq_data->hwirq] & IRQ_TYPE_LEVEL_MASK)
+		clk_disable(bank->clk);
+
  	gpiochip_unlock_as_irq(&bank->gpio_chip, irq_data->hwirq);
  }
 
#6 lore raw
In reply to Marek Vasut
Quoted 53 lines · Marek Vasut

The current EOI handler for LEVEL triggered interrupts calls clk_enable(), register IO, clk_disable(). The clock manipulation requires locking which happens with IRQs disabled in clk_enable_lock(). Instead of turning the clock on and off all the time, enable the clock in case LEVEL interrupt is requested and keep the clock enabled until all LEVEL interrupts are freed. The LEVEL interrupts are an exception on this platform and seldom used, so this does not affect the common case.

This simplifies the LEVEL interrupt handling considerably and also fixes the following splat found when using preempt-rt:

 ------------[ cut here ]------------
 WARNING: CPU: 0 PID: 0 at kernel/locking/rtmutex.c:2040 __rt_mutex_trylock+0x37/0x62
 Modules linked in:
 CPU: 0 PID: 0 Comm: swapper/0 Not tainted 5.10.109-rt65-stable-standard-00068-g6a5afc4b1217 #85
 Hardware name: STM32 (Device Tree Support)
 [<c010a45d>] (unwind_backtrace) from [<c010766f>] (show_stack+0xb/0xc)
 [<c010766f>] (show_stack) from [<c06353ab>] (dump_stack+0x6f/0x84)
 [<c06353ab>] (dump_stack) from [<c01145e3>] (__warn+0x7f/0xa4)
 [<c01145e3>] (__warn) from [<c063386f>] (warn_slowpath_fmt+0x3b/0x74)
 [<c063386f>] (warn_slowpath_fmt) from [<c063b43d>] (__rt_mutex_trylock+0x37/0x62)
 [<c063b43d>] (__rt_mutex_trylock) from [<c063c053>] (rt_spin_trylock+0x7/0x16)
 [<c063c053>] (rt_spin_trylock) from [<c036a2f3>] (clk_enable_lock+0xb/0x80)
 [<c036a2f3>] (clk_enable_lock) from [<c036ba69>] (clk_core_enable_lock+0x9/0x18)
 [<c036ba69>] (clk_core_enable_lock) from [<c034e9f3>] (stm32_gpio_get+0x11/0x24)
 [<c034e9f3>] (stm32_gpio_get) from [<c034ef43>] (stm32_gpio_irq_trigger+0x1f/0x48)
 [<c034ef43>] (stm32_gpio_irq_trigger) from [<c014aa53>] (handle_fasteoi_irq+0x71/0xa8)
 [<c014aa53>] (handle_fasteoi_irq) from [<c0147111>] (generic_handle_irq+0x19/0x22)
 [<c0147111>] (generic_handle_irq) from [<c014752d>] (__handle_domain_irq+0x55/0x64)
 [<c014752d>] (__handle_domain_irq) from [<c0346f13>] (gic_handle_irq+0x53/0x64)
 [<c0346f13>] (gic_handle_irq) from [<c0100ba5>] (__irq_svc+0x65/0xc0)
 Exception stack(0xc0e01f18 to 0xc0e01f60)
 1f00:                                                       0000300c 00000000
 1f20: 0000300c c010ff01 00000000 00000000 c0e00000 c0e07714 00000001 c0e01f78
 1f40: c0e07758 00000000 ef7cd0ff c0e01f68 c010554b c0105542 40000033 ffffffff
 [<c0100ba5>] (__irq_svc) from [<c0105542>] (arch_cpu_idle+0xc/0x1e)
 [<c0105542>] (arch_cpu_idle) from [<c063be95>] (default_idle_call+0x21/0x3c)
 [<c063be95>] (default_idle_call) from [<c01324f7>] (do_idle+0xe3/0x1e4)
 [<c01324f7>] (do_idle) from [<c01327b3>] (cpu_startup_entry+0x13/0x14)
 [<c01327b3>] (cpu_startup_entry) from [<c0a00c13>] (start_kernel+0x397/0x3d4)
 [<c0a00c13>] (start_kernel) from [<00000000>] (0x0)
 ---[ end trace 0000000000000002 ]---

Power consumption measured on STM32MP157C DHCOM SoM is not increased or is below noise threshold.

Fixes 47beed513a85b ("pinctrl: stm32: Add level interrupt support to gpio irq chip")
Cc Alexandre Torgue <[email protected]>
Cc Fabien Dessenne <[email protected]>
Cc Linus Walleij <[email protected]>
Cc Marc Zyngier <[email protected]>
Quoted 1 line

To: [email protected]

Patch applied to fixes, tweaked subject to pinctrl: stm32:

Yours,
Linus Walleij

_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel