Re: [PATCH v4 1/2] usb: ehci-exynos: Make provision for vdd regulators

From: Vivek Gautam
Date: Thu Jun 11 2015 - 14:44:15 EST


Hi,


On Mon, Jun 8, 2015 at 8:47 PM, Alan Stern <stern@xxxxxxxxxxxxxxxxxxx> wrote:
> On Mon, 8 Jun 2015, Vivek Gautam wrote:
>
>> Facilitate getting required 3.3V and 1.0V VDD supply for
>> EHCI controller on Exynos.
>>
>> For example, patches for regulators' nodes:
>> c8c253f ARM: dts: Add regulator entries to smdk5420
>> 275dcd2 ARM: dts: add max77686 pmic node for smdk5250,
>> enable only minimum number of regulators on smdk5250.
>>
>> So ensuring now that the controller driver requests the necessary
>> VDD regulators (if available, unless there are direct VDD rails),
>> and enable them so as to make them working on exynos systems.
>>
>> Signed-off-by: Vivek Gautam <gautam.vivek@xxxxxxxxxxx>
>
> Something about this looks a little fishy...

Sorry, i didn't get you.
This patch was initially posted by me sometime back.

>
>> @@ -170,7 +173,27 @@ static int exynos_ehci_probe(struct platform_device *pdev)
>>
>> err = exynos_ehci_get_phy(&pdev->dev, exynos_ehci);
>> if (err)
>> - goto fail_clk;
>> + goto fail_regulator1;
>> +
>> + exynos_ehci->vdd33 = devm_regulator_get_optional(&pdev->dev, "vdd33");
>
> Just before this region of code, there is:
>
> if (of_device_is_compatible(pdev->dev.of_node,
> "samsung,exynos5440-ehci"))
> goto skip_phy;
>
> If that "goto" is taken, exynos_ehci->vdd33 and ->vdd10 will be NULL,
> not an ERR_PTR code.

Right. This will hit NULL pointer dereferencing later in the code.

>
>> + if (!IS_ERR(exynos_ehci->vdd33)) {
>> + err = regulator_enable(exynos_ehci->vdd33);
>> + if (err) {
>> + dev_err(&pdev->dev,
>> + "Failed to enable 3.3V Vdd supply\n");
>> + goto fail_regulator1;
>> + }
>> + }

May be we can have something like this:

if (IS_ERR(exynos_ehci->vdd33)) {
exynos_ehci->vdd33 = NULL;
} else {
err = regulator_enable(exynos_ehci->vdd33);
if (err) {
dev_err(&pdev->dev,
"Failed to enable 3.3V Vdd supply\n");
goto fail_regulator1;
}
}

and later in the code check for NULL pointer before enabling the regulator.

>> +
>> + exynos_ehci->vdd10 = devm_regulator_get_optional(&pdev->dev, "vdd10");
>> + if (!IS_ERR(exynos_ehci->vdd10)) {
>> + err = regulator_enable(exynos_ehci->vdd10);
>> + if (err) {
>> + dev_err(&pdev->dev,
>> + "Failed to enable 1.0V Vdd supply\n");
>> + goto fail_regulator2;
>> + }
>> + }
>>
>> skip_phy:
>>
>> @@ -231,6 +254,12 @@ fail_add_hcd:
>> fail_io:
>> clk_disable_unprepare(exynos_ehci->clk);
>> fail_clk:
>> + if (!IS_ERR(exynos_ehci->vdd10))
>> + regulator_disable(exynos_ehci->vdd10);
>> +fail_regulator2:
>> + if (!IS_ERR(exynos_ehci->vdd33))
>> + regulator_disable(exynos_ehci->vdd33);
>
> Which means these regulator_disable() calls will crash when they
> dereference a NULL pointer.
>
> I think it would be simpler in the end to let a NULL pointer mean the
> regulator isn't present. If devm_regulator_get_optional() returns an
> error, convert it to NULL (or don't do the assignment to
> exynos_ehci->vdd?? in the first place).
>
> The same criticism applies to the other patch in this series.

Sure, i will amend both the patches.

Thanks !

--
Best Regards
Vivek Gautam
Samsung R&D Institute, Bangalore
India
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/