[PATCH v2] PCI: cadence: Fix runtime atomic count underflow.

[PATCH v2] PCI: cadence: Fix runtime atomic count underflow.

[Hans Zhang]

If the call to pci_host_probe() in cdns_pcie_host_setup()
fails, PM runtime count is decremented in the error path using
pm_runtime_put_sync().But the runtime count is not incremented by this
driver, but only by the callers (cdns_plat_pcie_probe/j721e_pcie_probe).
And the callers also decrement theruntime PM count in their error path.
So this leads to the below warning from the PM core:

runtime PM usage count underflow!

So fix it by getting rid of pm_runtime_put_sync() in the error path and
directly return the errno.

Fixes: 1b79c5284439 ("PCI: cadence: Add host driver for Cadence PCIe controller")

Signed-off-by: Hans Zhang <18255117159@163.com>
---
 drivers/pci/controller/cadence/pcie-cadence-host.c | 11 +----------
 1 file changed, 1 insertion(+), 10 deletions(-)

diff --git a/drivers/pci/controller/cadence/pcie-cadence-host.c b/drivers/pci/controller/cadence/pcie-cadence-host.c
index 8af95e9da7ce..741e10a575ec 100644
--- a/drivers/pci/controller/cadence/pcie-cadence-host.c
+++ b/drivers/pci/controller/cadence/pcie-cadence-host.c
@@ -570,14 +570,5 @@ int cdns_pcie_host_setup(struct cdns_pcie_rc *rc)
 	if (!bridge->ops)
 		bridge->ops = &cdns_pcie_host_ops;
 
-	ret = pci_host_probe(bridge);
-	if (ret < 0)
-		goto err_init;
-
-	return 0;
-
- err_init:
-	pm_runtime_put_sync(dev);
-
-	return ret;
+	return pci_host_probe(bridge);
 }

base-commit: a24588245776dafc227243a01bfbeb8a59bafba9
-- 
2.25.1

Re: [PATCH v2] PCI: cadence: Fix runtime atomic count underflow.

[Manivannan Sadhasivam]

On Sat, Apr 19, 2025 at 09:30:58PM +0800, Hans Zhang wrote:
> If the call to pci_host_probe() in cdns_pcie_host_setup()
> fails, PM runtime count is decremented in the error path using
> pm_runtime_put_sync().But the runtime count is not incremented by this
> driver, but only by the callers (cdns_plat_pcie_probe/j721e_pcie_probe).
> And the callers also decrement theruntime PM count in their error path.
> So this leads to the below warning from the PM core:
> 
> runtime PM usage count underflow!
> 
> So fix it by getting rid of pm_runtime_put_sync() in the error path and
> directly return the errno.
> 
> Fixes: 1b79c5284439 ("PCI: cadence: Add host driver for Cadence PCIe controller")
> 

This is not the correct Fixes commit. In fact it took me a while to find the
correct one. This exact same issue was already fixed by commit, 19abcd790b51
("PCI: cadence: Fix cdns_pcie_{host|ep}_setup() error path").

But then, this bug got reintroduced while fixing the merge conflict in:
49e427e6bdd1 ("Merge branch 'pci/host-probe-refactor'")

I will change the tag while applying.

- Mani

> Signed-off-by: Hans Zhang <18255117159@163.com>
> ---
>  drivers/pci/controller/cadence/pcie-cadence-host.c | 11 +----------
>  1 file changed, 1 insertion(+), 10 deletions(-)
> 
> diff --git a/drivers/pci/controller/cadence/pcie-cadence-host.c b/drivers/pci/controller/cadence/pcie-cadence-host.c
> index 8af95e9da7ce..741e10a575ec 100644
> --- a/drivers/pci/controller/cadence/pcie-cadence-host.c
> +++ b/drivers/pci/controller/cadence/pcie-cadence-host.c
> @@ -570,14 +570,5 @@ int cdns_pcie_host_setup(struct cdns_pcie_rc *rc)
>  	if (!bridge->ops)
>  		bridge->ops = &cdns_pcie_host_ops;
>  
> -	ret = pci_host_probe(bridge);
> -	if (ret < 0)
> -		goto err_init;
> -
> -	return 0;
> -
> - err_init:
> -	pm_runtime_put_sync(dev);
> -
> -	return ret;
> +	return pci_host_probe(bridge);
>  }
> 
> base-commit: a24588245776dafc227243a01bfbeb8a59bafba9
> -- 
> 2.25.1
> 

-- 
மணிவண்ணன் சதாசிவம்

Re: [PATCH v2] PCI: cadence: Fix runtime atomic count underflow.

[Hans Zhang]

On 2025/4/20 00:51, Manivannan Sadhasivam wrote:
> On Sat, Apr 19, 2025 at 09:30:58PM +0800, Hans Zhang wrote:
>> If the call to pci_host_probe() in cdns_pcie_host_setup()
>> fails, PM runtime count is decremented in the error path using
>> pm_runtime_put_sync().But the runtime count is not incremented by this
>> driver, but only by the callers (cdns_plat_pcie_probe/j721e_pcie_probe).
>> And the callers also decrement theruntime PM count in their error path.
>> So this leads to the below warning from the PM core:
>>
>> runtime PM usage count underflow!
>>
>> So fix it by getting rid of pm_runtime_put_sync() in the error path and
>> directly return the errno.
>>
>> Fixes: 1b79c5284439 ("PCI: cadence: Add host driver for Cadence PCIe controller")
>>
> 
> This is not the correct Fixes commit. In fact it took me a while to find the
> correct one. This exact same issue was already fixed by commit, 19abcd790b51
> ("PCI: cadence: Fix cdns_pcie_{host|ep}_setup() error path").
> 
> But then, this bug got reintroduced while fixing the merge conflict in:
> 49e427e6bdd1 ("Merge branch 'pci/host-probe-refactor'")
> 
> I will change the tag while applying.
> 

Hi Mani,

Thank you very much.

Best regards,
Hans

> - Mani
> 
>> Signed-off-by: Hans Zhang <18255117159@163.com>
>> ---
>>   drivers/pci/controller/cadence/pcie-cadence-host.c | 11 +----------
>>   1 file changed, 1 insertion(+), 10 deletions(-)
>>
>> diff --git a/drivers/pci/controller/cadence/pcie-cadence-host.c b/drivers/pci/controller/cadence/pcie-cadence-host.c
>> index 8af95e9da7ce..741e10a575ec 100644
>> --- a/drivers/pci/controller/cadence/pcie-cadence-host.c
>> +++ b/drivers/pci/controller/cadence/pcie-cadence-host.c
>> @@ -570,14 +570,5 @@ int cdns_pcie_host_setup(struct cdns_pcie_rc *rc)
>>   	if (!bridge->ops)
>>   		bridge->ops = &cdns_pcie_host_ops;
>>   
>> -	ret = pci_host_probe(bridge);
>> -	if (ret < 0)
>> -		goto err_init;
>> -
>> -	return 0;
>> -
>> - err_init:
>> -	pm_runtime_put_sync(dev);
>> -
>> -	return ret;
>> +	return pci_host_probe(bridge);
>>   }
>>
>> base-commit: a24588245776dafc227243a01bfbeb8a59bafba9
>> -- 
>> 2.25.1
>>
> 

Re: [PATCH v2] PCI: cadence: Fix runtime atomic count underflow.

[Manivannan Sadhasivam]

On Sat, 19 Apr 2025 21:30:58 +0800, Hans Zhang wrote:
> If the call to pci_host_probe() in cdns_pcie_host_setup()
> fails, PM runtime count is decremented in the error path using
> pm_runtime_put_sync().But the runtime count is not incremented by this
> driver, but only by the callers (cdns_plat_pcie_probe/j721e_pcie_probe).
> And the callers also decrement theruntime PM count in their error path.
> So this leads to the below warning from the PM core:
> 
> [...]

Applied, thanks!

[1/1] PCI: cadence: Fix runtime atomic count underflow.
      commit: f8015b6a0db95ce09aaacb236746f33b7a540a3e

Best regards,
-- 
Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>