Re: [PATCH v1 3/4] usb: dwc3: exynos: Use devm_regulator_bulk_get_enable() helper function

From: Christophe JAILLET
Date: Mon Apr 08 2024 - 16:35:36 EST

Le 08/04/2024 à 12:02, Anand Moon a écrit :
Hi Christophe,

On Fri, 5 Apr 2024 at 21:42, Christophe JAILLET
<christophe.jaillet@xxxxxxxxxx> wrote:

Le 05/04/2024 à 08:10, Anand Moon a écrit :
Hi Christophe, Krzysztof,

On Mon, 4 Mar 2024 at 17:16, Anand Moon <linux.amoon@xxxxxxxxx> wrote:

Hi Christophe,

On Sun, 3 Mar 2024 at 00:07, Christophe JAILLET
<christophe.jaillet@xxxxxxxxxx> wrote:

Le 02/03/2024 à 17:48, Anand Moon a écrit :
Hi Christophe,

On Sat, 2 Mar 2024 at 21:20, Christophe JAILLET
<christophe.jaillet@xxxxxxxxxx> wrote:

Le 01/03/2024 à 20:38, Anand Moon a écrit :
Use devm_regulator_bulk_get_enable() instead of open coded
'devm_regulator_get(), regulator_enable(), regulator_disable().

Signed-off-by: Anand Moon <linux.amoon@xxxxxxxxx>
---
drivers/usb/dwc3/dwc3-exynos.c | 49 +++-------------------------------
1 file changed, 4 insertions(+), 45 deletions(-)

diff --git a/drivers/usb/dwc3/dwc3-exynos.c b/drivers/usb/dwc3/dwc3-exynos.c
index 5d365ca51771..7c77f3c69825 100644
--- a/drivers/usb/dwc3/dwc3-exynos.c
+++ b/drivers/usb/dwc3/dwc3-exynos.c
@@ -32,9 +32,6 @@ struct dwc3_exynos {
struct clk *clks[DWC3_EXYNOS_MAX_CLOCKS];
int num_clks;
int suspend_clk_idx;
-
- struct regulator *vdd33;
- struct regulator *vdd10;
};

static int dwc3_exynos_probe(struct platform_device *pdev)
@@ -44,6 +41,7 @@ static int dwc3_exynos_probe(struct platform_device *pdev)
struct device_node *node = dev->of_node;
const struct dwc3_exynos_driverdata *driver_data;
int i, ret;
+ static const char * const regulators[] = { "vdd33", "vdd10" };

exynos = devm_kzalloc(dev, sizeof(*exynos), GFP_KERNEL);
if (!exynos)
@@ -78,27 +76,9 @@ static int dwc3_exynos_probe(struct platform_device *pdev)
if (exynos->suspend_clk_idx >= 0)
clk_prepare_enable(exynos->clks[exynos->suspend_clk_idx]);

- exynos->vdd33 = devm_regulator_get(dev, "vdd33");
- if (IS_ERR(exynos->vdd33)) {
- ret = PTR_ERR(exynos->vdd33);
- goto vdd33_err;
- }
- ret = regulator_enable(exynos->vdd33);
- if (ret) {
- dev_err(dev, "Failed to enable VDD33 supply\n");
- goto vdd33_err;
- }
-
- exynos->vdd10 = devm_regulator_get(dev, "vdd10");
- if (IS_ERR(exynos->vdd10)) {
- ret = PTR_ERR(exynos->vdd10);
- goto vdd10_err;
- }
- ret = regulator_enable(exynos->vdd10);
- if (ret) {
- dev_err(dev, "Failed to enable VDD10 supply\n");
- goto vdd10_err;
- }
+ ret = devm_regulator_bulk_get_enable(dev, ARRAY_SIZE(regulators), regulators);
+ if (ret)
+ return dev_err_probe(dev, ret, "Failed to enable regulators\n");

if (node) {
ret = of_platform_populate(node, NULL, NULL, dev);
@@ -115,10 +95,6 @@ static int dwc3_exynos_probe(struct platform_device *pdev)
return 0;

populate_err:
- regulator_disable(exynos->vdd10);
-vdd10_err:
- regulator_disable(exynos->vdd33);
-vdd33_err:
for (i = exynos->num_clks - 1; i >= 0; i--)
clk_disable_unprepare(exynos->clks[i]);

@@ -140,9 +116,6 @@ static void dwc3_exynos_remove(struct platform_device *pdev)

if (exynos->suspend_clk_idx >= 0)
clk_disable_unprepare(exynos->clks[exynos->suspend_clk_idx]);
-
- regulator_disable(exynos->vdd33);
- regulator_disable(exynos->vdd10);
}

static const struct dwc3_exynos_driverdata exynos5250_drvdata = {
@@ -196,9 +169,6 @@ static int dwc3_exynos_suspend(struct device *dev)
for (i = exynos->num_clks - 1; i >= 0; i--)
clk_disable_unprepare(exynos->clks[i]);

- regulator_disable(exynos->vdd33);
- regulator_disable(exynos->vdd10);

Hi,

Same here, I don't think that removing regulator_[en|dis]able from the
suspend and resume function is correct.

The goal is to stop some hardware when the system is suspended, in order
to save some power.
Ok,

Why did you removed it?

As per the description of the function devm_regulator_bulk_get_enable

* This helper function allows drivers to get several regulator
* consumers in one operation with management, the regulators will
* automatically be freed when the device is unbound. If any of the
* regulators cannot be acquired then any regulators that were
* allocated will be freed before returning to the caller.

The code in suspend/resume is not about freeing some resources. It is
about enabling/disabling some hardware to save some power.

Think to the probe/remove functions as the software in the kernel that
knows how to handle some hardawre, and the suspend/resume as the on/off
button to power-on and off the electrical chips.

When the system is suspended, the software is still around. But some
hardware can be set in a low consumption mode to save some power.

IMHO, part of the code you removed changed this behaviour and increase
the power consumption when the system is suspended.


You are correct, I have changed the regulator API from
devm_regulator_get_enable to devm_regulator_bulk_get_enable
which changes this behavior.
I will fix it in the next version.

CJ

I could not find any example in the kernel to support
devm_regulator_bulk_disable
but here is my modified file.

If you have any suggestions for this plz let me know.

I don't think that your approach is correct, and I don't think that the
proposed patch does what you expect it to do.

Calling a devm_ function in suspend/resume functions looks really
strange to me and is likely broken.

Especially here, devm_regulator_bulk_get_enable() in the resume function
allocates some memory that is not freed in
devm_regulator_bulk_disable(), because the API is not designed to work
like that. So this could generate a kind of memory leak.


*I think that the code is good enough as-is*, but if you really want to
change something, maybe:
- devm_regulator_get()+regulator_enable() in the probe could be
changed to devm_regulator_get_enable()
- the resume/suspend function should be left as-is with
regulator_disable()/regulator_ensable()
- remove regulator_disable() from the error handling path of the
probe and from the remove function.

I *think* it would work.

No devm_regulator_get_enable use the same logic as
devm_regulator_bulk_get_enable
to enable the regulator.

Yes, the logic is the same, but you get a pointer to the "struct regulator" which can be used to disable/enable in the suspend/resume functions.

With the bulk version, you can not do that.

See my first reply on your 3/4 patch.


[0] https://elixir.bootlin.com/linux/latest/source/drivers/regulator/devres.c#L126

So as of now I am dropping the changes on the regulator in this patch series.

I do agree that it is certainly the way to go here.

CJ


CJ

Thanks for your inputs.

Thanks

-Anand