[PATCH] PCI: sg2042: Fix a reference count issue in sg2042_pcie_remove()

[PATCH] PCI: sg2042: Fix a reference count issue in sg2042_pcie_remove()

[Christophe JAILLET]

devm_pm_runtime_enable() is used in the probe, so pm_runtime_disable()
should not be called explicitly in the remove function.

Fixes: 1c72774df028 ("PCI: sg2042: Add Sophgo SG2042 PCIe driver")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
---
Compile tested only
---
 drivers/pci/controller/cadence/pcie-sg2042.c | 3 ---
 1 file changed, 3 deletions(-)

diff --git a/drivers/pci/controller/cadence/pcie-sg2042.c b/drivers/pci/controller/cadence/pcie-sg2042.c
index a077b28d4894..0c50c74d03ee 100644
--- a/drivers/pci/controller/cadence/pcie-sg2042.c
+++ b/drivers/pci/controller/cadence/pcie-sg2042.c
@@ -74,15 +74,12 @@ static int sg2042_pcie_probe(struct platform_device *pdev)
 static void sg2042_pcie_remove(struct platform_device *pdev)
 {
 	struct cdns_pcie *pcie = platform_get_drvdata(pdev);
-	struct device *dev = &pdev->dev;
 	struct cdns_pcie_rc *rc;
 
 	rc = container_of(pcie, struct cdns_pcie_rc, pcie);
 	cdns_pcie_host_disable(rc);
 
 	cdns_pcie_disable_phy(pcie);
-
-	pm_runtime_disable(dev);
 }
 
 static int sg2042_pcie_suspend_noirq(struct device *dev)
-- 
2.51.0

Re: [PATCH] PCI: sg2042: Fix a reference count issue in sg2042_pcie_remove()

[Chen Wang]

On 9/30/2025 2:13 AM, Christophe JAILLET wrote:
> devm_pm_runtime_enable() is used in the probe, so pm_runtime_disable()
> should not be called explicitly in the remove function.
>
> Fixes: 1c72774df028 ("PCI: sg2042: Add Sophgo SG2042 PCIe driver")
> Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>

LGTM.

Acked-by: Chen Wang <unicorn_wang@outlook.com>

Tested-by: Chen Wang <unicorn_wang@outlook.com> # on Pioneerbox.

Thanks,

Chen

> ---
> Compile tested only
> ---
>   drivers/pci/controller/cadence/pcie-sg2042.c | 3 ---
>   1 file changed, 3 deletions(-)
>
> diff --git a/drivers/pci/controller/cadence/pcie-sg2042.c b/drivers/pci/controller/cadence/pcie-sg2042.c
> index a077b28d4894..0c50c74d03ee 100644
> --- a/drivers/pci/controller/cadence/pcie-sg2042.c
> +++ b/drivers/pci/controller/cadence/pcie-sg2042.c
> @@ -74,15 +74,12 @@ static int sg2042_pcie_probe(struct platform_device *pdev)
>   static void sg2042_pcie_remove(struct platform_device *pdev)
>   {
>   	struct cdns_pcie *pcie = platform_get_drvdata(pdev);
> -	struct device *dev = &pdev->dev;
>   	struct cdns_pcie_rc *rc;
>   
>   	rc = container_of(pcie, struct cdns_pcie_rc, pcie);
>   	cdns_pcie_host_disable(rc);
>   
>   	cdns_pcie_disable_phy(pcie);
> -
> -	pm_runtime_disable(dev);
>   }
>   
>   static int sg2042_pcie_suspend_noirq(struct device *dev)

Re: [PATCH] PCI: sg2042: Fix a reference count issue in sg2042_pcie_remove()

[Chen Wang]

Hi，Manivannan,

I see 6.18-rc1 is released. Could you please pick this fix for 6.18-rcX?

Thanks,

Chen

On 9/30/2025 8:43 AM, Chen Wang wrote:
>
> On 9/30/2025 2:13 AM, Christophe JAILLET wrote:
>> devm_pm_runtime_enable() is used in the probe, so pm_runtime_disable()
>> should not be called explicitly in the remove function.
>>
>> Fixes: 1c72774df028 ("PCI: sg2042: Add Sophgo SG2042 PCIe driver")
>> Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
>
> LGTM.
>
> Acked-by: Chen Wang <unicorn_wang@outlook.com>
>
> Tested-by: Chen Wang <unicorn_wang@outlook.com> # on Pioneerbox.
>
> Thanks,
>
> Chen
>
>> ---
>> Compile tested only
>> ---
>>   drivers/pci/controller/cadence/pcie-sg2042.c | 3 ---
>>   1 file changed, 3 deletions(-)
>>
>> diff --git a/drivers/pci/controller/cadence/pcie-sg2042.c 
>> b/drivers/pci/controller/cadence/pcie-sg2042.c
>> index a077b28d4894..0c50c74d03ee 100644
>> --- a/drivers/pci/controller/cadence/pcie-sg2042.c
>> +++ b/drivers/pci/controller/cadence/pcie-sg2042.c
>> @@ -74,15 +74,12 @@ static int sg2042_pcie_probe(struct 
>> platform_device *pdev)
>>   static void sg2042_pcie_remove(struct platform_device *pdev)
>>   {
>>       struct cdns_pcie *pcie = platform_get_drvdata(pdev);
>> -    struct device *dev = &pdev->dev;
>>       struct cdns_pcie_rc *rc;
>>         rc = container_of(pcie, struct cdns_pcie_rc, pcie);
>>       cdns_pcie_host_disable(rc);
>>         cdns_pcie_disable_phy(pcie);
>> -
>> -    pm_runtime_disable(dev);
>>   }
>>     static int sg2042_pcie_suspend_noirq(struct device *dev)

Re: [PATCH] PCI: sg2042: Fix a reference count issue in sg2042_pcie_remove()

[Bjorn Helgaas]

On Mon, Oct 13, 2025 at 10:31:22AM +0800, Chen Wang wrote:
> Hi，Manivannan,
> 
> I see 6.18-rc1 is released. Could you please pick this fix for 6.18-rcX?

Mani queued this for v6.19.  Is there a reason it should be in v6.18
instead?  We're after the v6.18 merge window now, so we only add
things to v6.18 if they fix a serious issue.

> On 9/30/2025 8:43 AM, Chen Wang wrote:
> > 
> > On 9/30/2025 2:13 AM, Christophe JAILLET wrote:
> > > devm_pm_runtime_enable() is used in the probe, so pm_runtime_disable()
> > > should not be called explicitly in the remove function.
> > > 
> > > Fixes: 1c72774df028 ("PCI: sg2042: Add Sophgo SG2042 PCIe driver")
> > > Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
> > 
> > LGTM.
> > 
> > Acked-by: Chen Wang <unicorn_wang@outlook.com>
> > 
> > Tested-by: Chen Wang <unicorn_wang@outlook.com> # on Pioneerbox.
> > 
> > Thanks,
> > 
> > Chen
> > 
> > > ---
> > > Compile tested only
> > > ---
> > >   drivers/pci/controller/cadence/pcie-sg2042.c | 3 ---
> > >   1 file changed, 3 deletions(-)
> > > 
> > > diff --git a/drivers/pci/controller/cadence/pcie-sg2042.c
> > > b/drivers/pci/controller/cadence/pcie-sg2042.c
> > > index a077b28d4894..0c50c74d03ee 100644
> > > --- a/drivers/pci/controller/cadence/pcie-sg2042.c
> > > +++ b/drivers/pci/controller/cadence/pcie-sg2042.c
> > > @@ -74,15 +74,12 @@ static int sg2042_pcie_probe(struct
> > > platform_device *pdev)
> > >   static void sg2042_pcie_remove(struct platform_device *pdev)
> > >   {
> > >       struct cdns_pcie *pcie = platform_get_drvdata(pdev);
> > > -    struct device *dev = &pdev->dev;
> > >       struct cdns_pcie_rc *rc;
> > >         rc = container_of(pcie, struct cdns_pcie_rc, pcie);
> > >       cdns_pcie_host_disable(rc);
> > >         cdns_pcie_disable_phy(pcie);
> > > -
> > > -    pm_runtime_disable(dev);
> > >   }
> > >     static int sg2042_pcie_suspend_noirq(struct device *dev)

Re: [PATCH] PCI: sg2042: Fix a reference count issue in sg2042_pcie_remove()

[Manivannan Sadhasivam]

On Mon, Oct 20, 2025 at 10:27:38AM -0500, Bjorn Helgaas wrote:
> On Mon, Oct 13, 2025 at 10:31:22AM +0800, Chen Wang wrote:
> > Hi，Manivannan,
> > 
> > I see 6.18-rc1 is released. Could you please pick this fix for 6.18-rcX?
> 
> Mani queued this for v6.19.  Is there a reason it should be in v6.18
> instead?  We're after the v6.18 merge window now, so we only add
> things to v6.18 if they fix a serious issue.
> 

The implication of calling pm_runtime_disable() manually in the remove() path
is, 'dev->power.disable_depth' will be incremented twice. When the driver gets
probed again, devm_pm_runtime_enable() will not enable runtime PM for the 'dev',
but just decrement 'dev->power.disable_depth'.

So this will result in an imbalanced runtime PM state. This is an issue, but
the severity is less since the driver is not itself making use of runtime PM.
So the driver should continue to work fine, but the runtime PM chain might be
broken.

- Mani

-- 
மணிவண்ணன் சதாசிவம்

Re: [PATCH] PCI: sg2042: Fix a reference count issue in sg2042_pcie_remove()

[Chen Wang]

Hi，Bjorn，

I see you raise some fixes PRs for PCI for 6.18,  can you please pick 
this one also?

Thanks,

Chen

On 10/13/2025 10:31 AM, Chen Wang wrote:
> Hi，Manivannan,
>
> I see 6.18-rc1 is released. Could you please pick this fix for 6.18-rcX?
>
> Thanks,
>
> Chen
>
> On 9/30/2025 8:43 AM, Chen Wang wrote:
>>
>> On 9/30/2025 2:13 AM, Christophe JAILLET wrote:
>>> devm_pm_runtime_enable() is used in the probe, so pm_runtime_disable()
>>> should not be called explicitly in the remove function.
>>>
>>> Fixes: 1c72774df028 ("PCI: sg2042: Add Sophgo SG2042 PCIe driver")
>>> Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
>>
>> LGTM.
>>
>> Acked-by: Chen Wang <unicorn_wang@outlook.com>
>>
>> Tested-by: Chen Wang <unicorn_wang@outlook.com> # on Pioneerbox.
>>
>> Thanks,
>>
>> Chen
>>
>>> ---
>>> Compile tested only
>>> ---
>>>   drivers/pci/controller/cadence/pcie-sg2042.c | 3 ---
>>>   1 file changed, 3 deletions(-)
>>>
>>> diff --git a/drivers/pci/controller/cadence/pcie-sg2042.c 
>>> b/drivers/pci/controller/cadence/pcie-sg2042.c
>>> index a077b28d4894..0c50c74d03ee 100644
>>> --- a/drivers/pci/controller/cadence/pcie-sg2042.c
>>> +++ b/drivers/pci/controller/cadence/pcie-sg2042.c
>>> @@ -74,15 +74,12 @@ static int sg2042_pcie_probe(struct 
>>> platform_device *pdev)
>>>   static void sg2042_pcie_remove(struct platform_device *pdev)
>>>   {
>>>       struct cdns_pcie *pcie = platform_get_drvdata(pdev);
>>> -    struct device *dev = &pdev->dev;
>>>       struct cdns_pcie_rc *rc;
>>>         rc = container_of(pcie, struct cdns_pcie_rc, pcie);
>>>       cdns_pcie_host_disable(rc);
>>>         cdns_pcie_disable_phy(pcie);
>>> -
>>> -    pm_runtime_disable(dev);
>>>   }
>>>     static int sg2042_pcie_suspend_noirq(struct device *dev)

Re: [PATCH] PCI: sg2042: Fix a reference count issue in sg2042_pcie_remove()

[Manivannan Sadhasivam]

On Mon, 29 Sep 2025 20:13:22 +0200, Christophe JAILLET wrote:
> devm_pm_runtime_enable() is used in the probe, so pm_runtime_disable()
> should not be called explicitly in the remove function.
> 
> 

Applied, thanks!

[1/1] PCI: sg2042: Fix a reference count issue in sg2042_pcie_remove()
      commit: 932ec9dff6da40382ee63049a11a6ff047bdc259

Best regards,
-- 
Manivannan Sadhasivam <mani@kernel.org>