usb: ehci-exynos: Use devm_clk_get_enabled() helpers

The devm_clk_get_enabled() helpers:
- call devm_clk_get()
- call clk_prepare_enable() and register what is needed in order to call clk_disable_unprepare() when needed, as a managed resource.

This simplifies the code and avoids the calls to clk_disable_unprepare().

While at it, use dev_err_probe consistently, and use its return value to return the error code.

Signed-off-by Anand Moon <[email protected]>

---
V2: drop the clk_disable_unprepare in suspend/resume functions
fix the usb_put_hcd return before the devm_clk_get_enabled

---
 drivers/usb/host/ehci-exynos.c | 19 +++++--------------
 1 file changed, 5 insertions(+), 14 deletions(-)

drivers/usb/host/ehci-exynos.c +0 -3

index f644b131cc0b..f00bfd0b13dc 100644
@@ -159,20 +159,15 @@ static int exynos_ehci_probe(struct platform_device *pdev)
 
 	err = exynos_ehci_get_phy(&pdev->dev, exynos_ehci);
 	if (err)
-		goto fail_clk;
-
-	exynos_ehci->clk = devm_clk_get(&pdev->dev, "usbhost");

Quoted 29 lines · Anand Moon

The devm_clk_get_enabled() helpers:
- call devm_clk_get()
- call clk_prepare_enable() and register what is needed in order to call clk_disable_unprepare() when needed, as a managed resource.

This simplifies the code and avoids the calls to clk_disable_unprepare().

While at it, use dev_err_probe consistently, and use its return value to return the error code.

Signed-off-by Anand Moon <[email protected]>

---
V2: drop the clk_disable_unprepare in suspend/resume functions
fix the usb_put_hcd return before the devm_clk_get_enabled

---
 drivers/usb/host/ehci-exynos.c | 19 +++++--------------
 1 file changed, 5 insertions(+), 14 deletions(-)

drivers/usb/host/ehci-exynos.c +0 -3

index f644b131cc0b..f00bfd0b13dc 100644
@@ -159,20 +159,15 @@ static int exynos_ehci_probe(struct platform_device *pdev)
 
      err = exynos_ehci_get_phy(&pdev->dev, exynos_ehci);
      if (err)
-             goto fail_clk;
-
-     exynos_ehci->clk = devm_clk_get(&pdev->dev, "usbhost");

drivers/usb/host/ehci-exynos.c static int exynos_ehci_probe(struct platform_device *pdev) · line 168

Quoted 10 lines

drivers/usb/host/ehci-exynos.c +5 -3

+             goto fail_io;
 
+     exynos_ehci->clk = devm_clk_get_enabled(&pdev->dev, "usbhost");
      if (IS_ERR(exynos_ehci->clk)) {
-             dev_err(&pdev->dev, "Failed to get usbhost clock\n");
-             err = PTR_ERR(exynos_ehci->clk);
-             goto fail_clk;
+             usb_put_hcd(hcd);
+             return dev_err_probe(&pdev->dev, PTR_ERR(exynos_ehci->clk),
+                               "Failed to get usbhost clock\n");

Why is this logic changed?

If you want to call dev_err_probe(), that's great, but do NOT mix it up with a commit that does something totally different.

When you say something like "while at it" in a changelog text, that is a HUGE hint that it needs to be a separate commit. Because of that reason alone, I can't take these, you know better :(

thanks,

greg k-h

Hi Greg,

Quoted 47 lines · Greg Kroah-Hartman

Quoted 29 lines · Anand Moon

The devm_clk_get_enabled() helpers:
- call devm_clk_get()
- call clk_prepare_enable() and register what is needed in order to call clk_disable_unprepare() when needed, as a managed resource.

This simplifies the code and avoids the calls to clk_disable_unprepare().

While at it, use dev_err_probe consistently, and use its return value to return the error code.

Signed-off-by Anand Moon <[email protected]>

---
V2: drop the clk_disable_unprepare in suspend/resume functions
fix the usb_put_hcd return before the devm_clk_get_enabled

---
 drivers/usb/host/ehci-exynos.c | 19 +++++--------------
 1 file changed, 5 insertions(+), 14 deletions(-)

drivers/usb/host/ehci-exynos.c +0 -3

index f644b131cc0b..f00bfd0b13dc 100644
@@ -159,20 +159,15 @@ static int exynos_ehci_probe(struct platform_device *pdev)
 
      err = exynos_ehci_get_phy(&pdev->dev, exynos_ehci);
      if (err)
-             goto fail_clk;
-
-     exynos_ehci->clk = devm_clk_get(&pdev->dev, "usbhost");

drivers/usb/host/ehci-exynos.c static int exynos_ehci_probe(struct platform_device *pdev) · line 168

Quoted 10 lines

drivers/usb/host/ehci-exynos.c +5 -3

+             goto fail_io;
 
+     exynos_ehci->clk = devm_clk_get_enabled(&pdev->dev, "usbhost");
      if (IS_ERR(exynos_ehci->clk)) {
-             dev_err(&pdev->dev, "Failed to get usbhost clock\n");
-             err = PTR_ERR(exynos_ehci->clk);
-             goto fail_clk;
+             usb_put_hcd(hcd);
+             return dev_err_probe(&pdev->dev, PTR_ERR(exynos_ehci->clk),
+                               "Failed to get usbhost clock\n");

Why is this logic changed?

If you want to call dev_err_probe(), that's great, but do NOT mix it up with a commit that does something totally different.

Quoted 6 lines

When you say something like "while at it" in a changelog text, that is a HUGE hint that it needs to be a separate commit. Because of that reason alone, I can't take these, you know better :(

thanks,

Ok, I will improve the commit message relevant to the code changes.

Quoted 55 lines · Anand Moon

Hi Greg,

Quoted 51 lines · Greg Kroah-Hartman

Quoted 29 lines · Anand Moon

The devm_clk_get_enabled() helpers:
- call devm_clk_get()
- call clk_prepare_enable() and register what is needed in order to call clk_disable_unprepare() when needed, as a managed resource.

This simplifies the code and avoids the calls to clk_disable_unprepare().

While at it, use dev_err_probe consistently, and use its return value to return the error code.

Signed-off-by Anand Moon <[email protected]>

---
V2: drop the clk_disable_unprepare in suspend/resume functions
fix the usb_put_hcd return before the devm_clk_get_enabled

---
 drivers/usb/host/ehci-exynos.c | 19 +++++--------------
 1 file changed, 5 insertions(+), 14 deletions(-)

drivers/usb/host/ehci-exynos.c +0 -3

index f644b131cc0b..f00bfd0b13dc 100644
@@ -159,20 +159,15 @@ static int exynos_ehci_probe(struct platform_device *pdev)
 
      err = exynos_ehci_get_phy(&pdev->dev, exynos_ehci);
      if (err)
-             goto fail_clk;
-
-     exynos_ehci->clk = devm_clk_get(&pdev->dev, "usbhost");

drivers/usb/host/ehci-exynos.c static int exynos_ehci_probe(struct platform_device *pdev) · line 168

Quoted 10 lines

drivers/usb/host/ehci-exynos.c +5 -3

+             goto fail_io;
 
+     exynos_ehci->clk = devm_clk_get_enabled(&pdev->dev, "usbhost");
      if (IS_ERR(exynos_ehci->clk)) {
-             dev_err(&pdev->dev, "Failed to get usbhost clock\n");
-             err = PTR_ERR(exynos_ehci->clk);
-             goto fail_clk;
+             usb_put_hcd(hcd);
+             return dev_err_probe(&pdev->dev, PTR_ERR(exynos_ehci->clk),
+                               "Failed to get usbhost clock\n");

Why is this logic changed?

If you want to call dev_err_probe(), that's great, but do NOT mix it up with a commit that does something totally different.

When you say something like "while at it" in a changelog text, that is a HUGE hint that it needs to be a separate commit. Because of that reason alone, I can't take these, you know better :(

Quoted 4 lines

Quoted 2 lines

thanks,

Ok, I will improve the commit message relevant to the code changes.

Please read Greg's message one more time. He did not propose to fix commit msg, right?

Use the new PM macros for the suspend and resume functions to be automatically dropped by the compiler when CONFIG_PM are disabled, without having to use #ifdef guards. If CONFIG_PM unused, they will simply be discarded by the compiler.

Signed-off-by Anand Moon <[email protected]>

---
v2: add __maybe_unused to suspend/resume functions in case CONFIG_PM is disabled.
dropped RUNTIME_PM_OPS
---

 drivers/usb/host/ehci-exynos.c | 11 +++--------
 1 file changed, 3 insertions(+), 8 deletions(-)

drivers/usb/host/ehci-exynos.c +3 -8

index f00bfd0b13dc..4cfde1af32be 100644
@@ -236,8 +236,7 @@ static void exynos_ehci_remove(struct platform_device *pdev)
 	usb_put_hcd(hcd);
 }
 
-#ifdef CONFIG_PM
-static int exynos_ehci_suspend(struct device *dev)
+static int __maybe_unused exynos_ehci_suspend(struct device *dev)
 {
 	struct usb_hcd *hcd = dev_get_drvdata(dev);
 	struct exynos_ehci_hcd *exynos_ehci = to_exynos_ehci(hcd);

@@ -256,7 +255,7 @@ static int exynos_ehci_suspend(struct device *dev)
 	return rc;
 }
 
-static int exynos_ehci_resume(struct device *dev)
+static int __maybe_unused exynos_ehci_resume(struct device *dev)
 {
 	struct usb_hcd *hcd = dev_get_drvdata(dev);
 	struct exynos_ehci_hcd *exynos_ehci = to_exynos_ehci(hcd);

@@ -279,10 +278,6 @@ static int exynos_ehci_resume(struct device *dev)
 	ehci_resume(hcd, false);
 	return 0;
 }
-#else
-#define exynos_ehci_suspend	NULL
-#define exynos_ehci_resume	NULL
-#endif
 
 static const struct dev_pm_ops exynos_ehci_pm_ops = {
 	.suspend	= exynos_ehci_suspend,

@@ -303,7 +298,7 @@ static struct platform_driver exynos_ehci_driver = {
 	.shutdown	= usb_hcd_platform_shutdown,
 	.driver = {
 		.name	= "exynos-ehci",
-		.pm	= &exynos_ehci_pm_ops,
+		.pm	= pm_ptr(&exynos_ehci_pm_ops),
 		.of_match_table = of_match_ptr(exynos_ehci_match),
 	}
 };

Signature · 1 line

2.44.0

The devm_clk_get_enabled() helpers:
- call devm_clk_get()
- call clk_prepare_enable() and register what is needed in order to call clk_disable_unprepare() when needed, as a managed resource.

This simplifies the code and avoids the calls to clk_disable_unprepare().

While at it, use dev_err_probe consistently, and use its return value to return the error code.

Signed-off-by Anand Moon <[email protected]>

---
v2: new changes in this series.

---
 drivers/usb/host/ohci-exynos.c | 19 +++++--------------
 1 file changed, 5 insertions(+), 14 deletions(-)

drivers/usb/host/ohci-exynos.c +0 -3

index 20e26a474591..85d04ae0ae40 100644
@@ -135,20 +135,15 @@ static int exynos_ohci_probe(struct platform_device *pdev)
 
 	err = exynos_ohci_get_phy(&pdev->dev, exynos_ohci);
 	if (err)
-		goto fail_clk;
-
-	exynos_ohci->clk = devm_clk_get(&pdev->dev, "usbhost");

Quoted 30 lines · Anand Moon

The devm_clk_get_enabled() helpers:
- call devm_clk_get()
- call clk_prepare_enable() and register what is needed in order to call clk_disable_unprepare() when needed, as a managed resource.

This simplifies the code and avoids the calls to clk_disable_unprepare().

While at it, use dev_err_probe consistently, and use its return value to return the error code.

Signed-off-by Anand Moon <[email protected]>

---
v2: new changes in this series.

---
 drivers/usb/host/ohci-exynos.c | 19 +++++--------------
 1 file changed, 5 insertions(+), 14 deletions(-)

drivers/usb/host/ohci-exynos.c +1 -3

index 20e26a474591..85d04ae0ae40 100644
@@ -135,20 +135,15 @@ static int exynos_ohci_probe(struct platform_device *pdev)
 
      err = exynos_ohci_get_phy(&pdev->dev, exynos_ohci);
      if (err)
-             goto fail_clk;
-
-     exynos_ohci->clk = devm_clk_get(&pdev->dev, "usbhost");
+             goto fail_io;
 

drivers/usb/host/ohci-exynos.c static int exynos_ohci_probe(struct platform_device *pdev) · line 144

Quoted 8 lines

drivers/usb/host/ohci-exynos.c +4 -3

+     exynos_ohci->clk = devm_clk_get_enabled(&pdev->dev, "usbhost");
      if (IS_ERR(exynos_ohci->clk)) {
-             dev_err(&pdev->dev, "Failed to get usbhost clock\n");
-             err = PTR_ERR(exynos_ohci->clk);
-             goto fail_clk;
+             usb_put_hcd(hcd);
+             return dev_err_probe(&pdev->dev, PTR_ERR(exynos_ohci->clk),
+                             "Failed to get usbhost clock\n");

Why do you introduce entirely parallel exit paths? There is already single error handling part with labels. Use that.

Best regards,
Krzysztof

Use the new PM macros for the suspend and resume functions to be automatically dropped by the compiler when CONFIG_PM are disabled, without having to use #ifdef guards. If CONFIG_PM unused, they will simply be discarded by the compiler.

Signed-off-by Anand Moon <[email protected]>

---
v2: new changes in this series.
---

 drivers/usb/host/ohci-exynos.c | 11 +++--------
 1 file changed, 3 insertions(+), 8 deletions(-)

drivers/usb/host/ohci-exynos.c +3 -8

index 85d04ae0ae40..3e647e0b341d 100644
@@ -212,8 +212,7 @@ static void exynos_ohci_shutdown(struct platform_device *pdev)
 		hcd->driver->shutdown(hcd);
 }
 
-#ifdef CONFIG_PM
-static int exynos_ohci_suspend(struct device *dev)
+static int __maybe_unused exynos_ohci_suspend(struct device *dev)
 {
 	struct usb_hcd *hcd = dev_get_drvdata(dev);
 	struct exynos_ohci_hcd *exynos_ohci = to_exynos_ohci(hcd);

@@ -230,7 +229,7 @@ static int exynos_ohci_suspend(struct device *dev)
 	return 0;
 }
 
-static int exynos_ohci_resume(struct device *dev)
+static int __maybe_unused exynos_ohci_resume(struct device *dev)
 {
 	struct usb_hcd *hcd			= dev_get_drvdata(dev);
 	struct exynos_ohci_hcd *exynos_ohci	= to_exynos_ohci(hcd);

@@ -249,10 +248,6 @@ static int exynos_ohci_resume(struct device *dev)
 
 	return 0;
 }
-#else
-#define exynos_ohci_suspend	NULL
-#define exynos_ohci_resume	NULL
-#endif
 
 static const struct ohci_driver_overrides exynos_overrides __initconst = {
 	.extra_priv_size =	sizeof(struct exynos_ohci_hcd),

@@ -277,7 +272,7 @@ static struct platform_driver exynos_ohci_driver = {
 	.shutdown	= exynos_ohci_shutdown,
 	.driver = {
 		.name	= "exynos-ohci",
-		.pm	= &exynos_ohci_pm_ops,
+		.pm	= pm_ptr(&exynos_ohci_pm_ops),
 		.of_match_table	= of_match_ptr(exynos_ohci_match),
 	}
 };

Signature · 1 line

2.44.0

Use devm_regulator_bulk_get_enable() instead of open coded 'devm_regulator_get(), regulator_enable(), regulator_disable().

Signed-off-by Anand Moon <[email protected]>

---
V2: no changes, did not find any regression in pm suspend/resume.

Quoted 2 lines · Anand Moon

Use devm_regulator_bulk_get_enable() instead of open coded 'devm_regulator_get(), regulator_enable(), regulator_disable().

I fail to see how did you replace open-coded suspend/resume paths.

Quoted 4 lines

Signed-off-by Anand Moon <[email protected]>

---
V2: no changes, did not find any regression in pm suspend/resume.

No, that's not equivalent code. No explanation in commit msg.

You already got comments on this and nothing improved. You just entirely ignored received comments. That's not how it works.

I don't think you understand the code and Linux driver model. This patch repeats several previous attempts with similar issues: no logic behind a change.

NAK.

Best regards,
Krzysztof

Hi Krzysztof,

Quoted 21 lines · Krzysztof Kozlowski

Quoted 2 lines · Anand Moon

Use devm_regulator_bulk_get_enable() instead of open coded 'devm_regulator_get(), regulator_enable(), regulator_disable().

I fail to see how did you replace open-coded suspend/resume paths.

Quoted 4 lines

Signed-off-by Anand Moon <[email protected]>

---
V2: no changes, did not find any regression in pm suspend/resume.

No, that's not equivalent code. No explanation in commit msg.

You already got comments on this and nothing improved. You just entirely ignored received comments. That's not how it works.

I don't think you understand the code and Linux driver model. This patch repeats several previous attempts with similar issues: no logic behind a change.

Quoted 1 line

NAK.

devm_regulator_get_enable and devm_regulator_bulk_get_enable
both remove the dependency from the driver to handle the regulator_enabled and regulator_disabled. ie this removes the regulator from the driver structure.

Since these functions set devm_add_action to disable the regulator when the resource is not used.

     ret = devm_add_action(dev, devm_regulator_bulk_disable, devres);
     if (!ret)
               return 0;

Use the new PM macros for the suspend and resume functions to be automatically dropped by the compiler when CONFIG_PM_SLEEP are disabled, without having to use #ifdef guards. If CONFIG_PM_SLEEP unused, they will simply be discarded by the compiler.

Signed-off-by Anand Moon <[email protected]>

---
v2: add __maybe_unused to suspend/resume functions in case CONFIG_PM is
disabled.

---
 drivers/usb/dwc3/dwc3-exynos.c | 12 +++---------
 1 file changed, 3 insertions(+), 9 deletions(-)

Quoted 5 lines · Anand Moon

Use the new PM macros for the suspend and resume functions to be automatically dropped by the compiler when CONFIG_PM_SLEEP are disabled, without having to use #ifdef guards. If CONFIG_PM_SLEEP unused, they will simply be discarded by the compiler.

Quoted 4 lines

Signed-off-by Anand Moon <[email protected]>

---
v2: add __maybe_unused to suspend/resume functions in case CONFIG_PM is
disabled.

The compiler discards the code, yet we still need __maybe_unused?

BR,
Thinh

Hi Thinh,

Quoted 7 lines · Thinh Nguyen

Quoted 5 lines · Anand Moon

Use the new PM macros for the suspend and resume functions to be automatically dropped by the compiler when CONFIG_PM_SLEEP are disabled, without having to use #ifdef guards. If CONFIG_PM_SLEEP unused, they will simply be discarded by the compiler.

Quoted 7 lines

Quoted 4 lines

Signed-off-by Anand Moon <[email protected]>

---
v2: add __maybe_unused to suspend/resume functions in case CONFIG_PM is
disabled.

The compiler discards the code, yet we still need __maybe_unused?

Earlier version had not added this since but I removed the guard.CONFIG_PM_SLEEP.
added __maybe_unused just to safeguard the function.

I have tried to build with config by disabling CONFIG_PM and CONFIG_PM_SLEEP but could get the warning compilation by adding flag W=1 -Werror=unused-function.

But since these CONFIG_PM and CONFIG_PM_SLEEP cannot be disabled, I am not getting any warning related to these functions.

Do you want me to remove __maybe_unused ?

Quoted 25 lines · Anand Moon

Hi Thinh,

Quoted 7 lines · Thinh Nguyen

Quoted 5 lines · Anand Moon

Use the new PM macros for the suspend and resume functions to be automatically dropped by the compiler when CONFIG_PM_SLEEP are disabled, without having to use #ifdef guards. If CONFIG_PM_SLEEP unused, they will simply be discarded by the compiler.

Quoted 7 lines

Quoted 4 lines

Signed-off-by Anand Moon <[email protected]>

---
v2: add __maybe_unused to suspend/resume functions in case CONFIG_PM is
disabled.

The compiler discards the code, yet we still need __maybe_unused?

Earlier version had not added this since but I removed the guard.CONFIG_PM_SLEEP.
added __maybe_unused just to safeguard the function.

I have tried to build with config by disabling CONFIG_PM and CONFIG_PM_SLEEP but could get the warning compilation by adding flag W=1 -Werror=unused-function.

<snip>

Quoted 4 lines

But since these CONFIG_PM and CONFIG_PM_SLEEP cannot be disabled, I am not getting any warning related to these functions.

Quoted 2 lines

Do you want me to remove __maybe_unused ?

The warning was there as expected. You should to use it along with DEFINE_SIMPLE_DEV_PM_OPS(). Let me know if you still see the same warning.

BR,
Thinh

Hi Thinh,

Quoted 34 lines · Thinh Nguyen

Quoted 25 lines · Anand Moon

Hi Thinh,

Quoted 7 lines · Thinh Nguyen

Quoted 5 lines · Anand Moon

Use the new PM macros for the suspend and resume functions to be automatically dropped by the compiler when CONFIG_PM_SLEEP are disabled, without having to use #ifdef guards. If CONFIG_PM_SLEEP unused, they will simply be discarded by the compiler.

Quoted 7 lines

Quoted 4 lines

Signed-off-by Anand Moon <[email protected]>

---
v2: add __maybe_unused to suspend/resume functions in case CONFIG_PM is
disabled.

The compiler discards the code, yet we still need __maybe_unused?

Earlier version had not added this since but I removed the guard.CONFIG_PM_SLEEP.
added __maybe_unused just to safeguard the function.

I have tried to build with config by disabling CONFIG_PM and CONFIG_PM_SLEEP but could get the warning compilation by adding flag W=1 -Werror=unused-function.

<snip>

Quoted 4 lines

But since these CONFIG_PM and CONFIG_PM_SLEEP cannot be disabled, I am not getting any warning related to these functions.

Quoted 7 lines

Quoted 2 lines

Do you want me to remove __maybe_unused ?

The warning was there as expected. You should to use it along with DEFINE_SIMPLE_DEV_PM_OPS(). Let me know if you still see the same warning.

But the warnings are related to the following macros

#define PTR_IF(cond, ptr) ((cond) ? (ptr) : NULL)

#define pm_ptr(_ptr) PTR_IF(IS_ENABLED(CONFIG_PM), (_ptr)) #define pm_sleep_ptr(_ptr) PTR_IF(IS_ENABLED(CONFIG_PM_SLEEP), (_ptr))

So if we can disable CONFIG_PM and CONFIG_PM_SLEEP options
the relevant function with the above macro will be set to NULL.
in this case, the compiler will discard the function in SET_SYSTEM_SLEEP_PM_OPS

Quoted 47 lines · Anand Moon

Hi Thinh,

Quoted 34 lines · Thinh Nguyen

Quoted 25 lines · Anand Moon

Hi Thinh,

Quoted 7 lines · Thinh Nguyen

Quoted 5 lines · Anand Moon

Use the new PM macros for the suspend and resume functions to be automatically dropped by the compiler when CONFIG_PM_SLEEP are disabled, without having to use #ifdef guards. If CONFIG_PM_SLEEP unused, they will simply be discarded by the compiler.

Quoted 7 lines

Quoted 4 lines

Signed-off-by Anand Moon <[email protected]>

---
v2: add __maybe_unused to suspend/resume functions in case CONFIG_PM is
disabled.

The compiler discards the code, yet we still need __maybe_unused?

Earlier version had not added this since but I removed the guard.CONFIG_PM_SLEEP.
added __maybe_unused just to safeguard the function.

I have tried to build with config by disabling CONFIG_PM and CONFIG_PM_SLEEP but could get the warning compilation by adding flag W=1 -Werror=unused-function.

<snip>

Quoted 4 lines

But since these CONFIG_PM and CONFIG_PM_SLEEP cannot be disabled, I am not getting any warning related to these functions.

Quoted 7 lines

Quoted 2 lines

Do you want me to remove __maybe_unused ?

The warning was there as expected. You should to use it along with DEFINE_SIMPLE_DEV_PM_OPS(). Let me know if you still see the same warning.

But the warnings are related to the following macros

Quoted 9 lines

#define PTR_IF(cond, ptr) ((cond) ? (ptr) : NULL)

#define pm_ptr(_ptr) PTR_IF(IS_ENABLED(CONFIG_PM), (_ptr)) #define pm_sleep_ptr(_ptr) PTR_IF(IS_ENABLED(CONFIG_PM_SLEEP), (_ptr))

So if we can disable CONFIG_PM and CONFIG_PM_SLEEP options
the relevant function with the above macro will be set to NULL.
in this case, the compiler will discard the function in SET_SYSTEM_SLEEP_PM_OPS

There are differences if you use SET_SYSTEM_SLEEP_PM_OPS vs the new macro. It's noted in this commit:

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=1a3c7bb088266fa2db017be299f91f1c1894c857

Please try it out.

Thanks,
Thinh