[PATCH] Bug fix for PME interrupt handler, add Root Status check

[PATCH] Bug fix for PME interrupt handler, add Root Status check

[Qiang]

PCIe PME and hot plug share same interrupt number. In some special case,
Link down event cause hot plug interrupt, devices is not disconnected,
But read config will return 0xff.

In that case, PME work function will run and not return Because
Root Status PME bit always 1 and can not be cleared.

This patch add Root Status check in PME interrupt handler,
Just do same as pciehp isr Slot status check.

Signed-off-by: Qiang Zheng <zhengqiang10@huawei.com>
---
 drivers/pci/pcie/pme.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/pci/pcie/pme.c b/drivers/pci/pcie/pme.c
index fafdb16..67144f5 100644
--- a/drivers/pci/pcie/pme.c
+++ b/drivers/pci/pcie/pme.c
@@ -273,7 +273,7 @@ static irqreturn_t pcie_pme_irq(int irq, void *context)
 	spin_lock_irqsave(&data->lock, flags);
 	pcie_capability_read_dword(port, PCI_EXP_RTSTA, &rtsta);
 
-	if (!(rtsta & PCI_EXP_RTSTA_PME)) {
+	if (rtsta == (u32)(~0) || !(rtsta & PCI_EXP_RTSTA_PME)) {
 		spin_unlock_irqrestore(&data->lock, flags);
 		return IRQ_NONE;
 	}
-- 
1.9.1

Re: [PATCH] Bug fix for PME interrupt handler, add Root Status check

[Rafael J. Wysocki]

On Thu, Sep 28, 2017 at 5:54 AM, Qiang <zhengqiang10@huawei.com> wrote:
> PCIe PME and hot plug share same interrupt number. In some special case,
> Link down event cause hot plug interrupt, devices is not disconnected,
> But read config will return 0xff.
>
> In that case, PME work function will run and not return Because
> Root Status PME bit always 1 and can not be cleared.
>
> This patch add Root Status check in PME interrupt handler,
> Just do same as pciehp isr Slot status check.
>
> Signed-off-by: Qiang Zheng <zhengqiang10@huawei.com>
> ---
>  drivers/pci/pcie/pme.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/pci/pcie/pme.c b/drivers/pci/pcie/pme.c
> index fafdb16..67144f5 100644
> --- a/drivers/pci/pcie/pme.c
> +++ b/drivers/pci/pcie/pme.c
> @@ -273,7 +273,7 @@ static irqreturn_t pcie_pme_irq(int irq, void *context)
>         spin_lock_irqsave(&data->lock, flags);
>         pcie_capability_read_dword(port, PCI_EXP_RTSTA, &rtsta);
>
> -       if (!(rtsta & PCI_EXP_RTSTA_PME)) {
> +       if (rtsta == (u32)(~0) || !(rtsta & PCI_EXP_RTSTA_PME)) {
>                 spin_unlock_irqrestore(&data->lock, flags);
>                 return IRQ_NONE;
>         }

OK, but isn't there a symbol for all ones in u32?  And if there isn't,
what about adding it?

Thanks,
Rafael

Re: [PATCH] Bug fix for PME interrupt handler, add Root Status check

[Zhengqiang]

Hi,Rafael, thanks for you reply.
please see my comment.

On 2017/9/29 7:13, Rafael J. Wysocki wrote:
> On Thu, Sep 28, 2017 at 5:54 AM, Qiang <zhengqiang10@huawei.com> wrote:
>> PCIe PME and hot plug share same interrupt number. In some special case,
>> Link down event cause hot plug interrupt, devices is not disconnected,
>> But read config will return 0xff.
>>
>> In that case, PME work function will run and not return Because
>> Root Status PME bit always 1 and can not be cleared.
>>
>> This patch add Root Status check in PME interrupt handler,
>> Just do same as pciehp isr Slot status check.
>>
>> Signed-off-by: Qiang Zheng <zhengqiang10@huawei.com>
>> ---
>>  drivers/pci/pcie/pme.c | 2 +-
>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/drivers/pci/pcie/pme.c b/drivers/pci/pcie/pme.c
>> index fafdb16..67144f5 100644
>> --- a/drivers/pci/pcie/pme.c
>> +++ b/drivers/pci/pcie/pme.c
>> @@ -273,7 +273,7 @@ static irqreturn_t pcie_pme_irq(int irq, void *context)
>>         spin_lock_irqsave(&data->lock, flags);
>>         pcie_capability_read_dword(port, PCI_EXP_RTSTA, &rtsta);
>>
>> -       if (!(rtsta & PCI_EXP_RTSTA_PME)) {
>> +       if (rtsta == (u32)(~0) || !(rtsta & PCI_EXP_RTSTA_PME)) {
>>                 spin_unlock_irqrestore(&data->lock, flags);
>>                 return IRQ_NONE;
>>         }
> 
> OK, but isn't there a symbol for all ones in u32?  And if there isn't,
> what about adding it?
> 
Here is Root Status register definitions.

Bit Location    Register Description        					Attributes
15:0            PME Requester ID – 						RO
		This field indicates the PCI Requester ID of the
		last PME Requester. This field is only valid
		when the PME Status bit is Set.

16              PME Status – This bit RW1C					RW1C
		indicates that PME was asserted by the PME Requester
		indicated in the PME Requester ID field. Subsequent PMEs
		are kept pending until the status register is cleared by
		software by writing a 1b.
		Default value of this bit is 0b.

17              PME Pending – 							RO
		This bit indicates that another PME is pending 		
		when the PME Status bit is Set. When the PME Status bit is
		cleared by software; the PME is delivered by hardware by
		setting the PME Status bit again and updating the PME
		Requester ID field appropriately. The PME Pending bit is
		cleared by hardware if no more PMEs are pending.
31:18		RsvdZ

Bit[31:18] is RsvdZ, read only and must return zero when read.

So, i guess there is no a symbol for all ones in u32.
If read Root Status return 0xff, perhaps RC is disconnected or accessed timeout.

> Thanks,
> Rafael
> 
> .
> 
Thanks

Re: [PATCH] Bug fix for PME interrupt handler, add Root Status check

[Rafael J. Wysocki]

On Fri, Sep 29, 2017 at 8:10 AM, Zhengqiang <zhengqiang10@huawei.com> wrote:
> Hi,Rafael, thanks for you reply.
> please see my comment.
>
> On 2017/9/29 7:13, Rafael J. Wysocki wrote:
>> On Thu, Sep 28, 2017 at 5:54 AM, Qiang <zhengqiang10@huawei.com> wrote:
>>> PCIe PME and hot plug share same interrupt number. In some special case,
>>> Link down event cause hot plug interrupt, devices is not disconnected,
>>> But read config will return 0xff.
>>>
>>> In that case, PME work function will run and not return Because
>>> Root Status PME bit always 1 and can not be cleared.
>>>
>>> This patch add Root Status check in PME interrupt handler,
>>> Just do same as pciehp isr Slot status check.
>>>
>>> Signed-off-by: Qiang Zheng <zhengqiang10@huawei.com>
>>> ---
>>>  drivers/pci/pcie/pme.c | 2 +-
>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>
>>> diff --git a/drivers/pci/pcie/pme.c b/drivers/pci/pcie/pme.c
>>> index fafdb16..67144f5 100644
>>> --- a/drivers/pci/pcie/pme.c
>>> +++ b/drivers/pci/pcie/pme.c
>>> @@ -273,7 +273,7 @@ static irqreturn_t pcie_pme_irq(int irq, void *context)
>>>         spin_lock_irqsave(&data->lock, flags);
>>>         pcie_capability_read_dword(port, PCI_EXP_RTSTA, &rtsta);
>>>
>>> -       if (!(rtsta & PCI_EXP_RTSTA_PME)) {
>>> +       if (rtsta == (u32)(~0) || !(rtsta & PCI_EXP_RTSTA_PME)) {
>>>                 spin_unlock_irqrestore(&data->lock, flags);
>>>                 return IRQ_NONE;
>>>         }
>>
>> OK, but isn't there a symbol for all ones in u32?  And if there isn't,
>> what about adding it?
>>
> Here is Root Status register definitions.

Well, that's not what I meant.

You have "rtsta == (u32)(~0)" in the patch above, but you could do
"rtsta == U32_MAX" instead.

Thanks,
Rafael

Re: [PATCH] Bug fix for PME interrupt handler, add Root Status check

[Zhengqiang]

On 2017/9/29 21:12, Rafael J. Wysocki wrote:
> On Fri, Sep 29, 2017 at 8:10 AM, Zhengqiang <zhengqiang10@huawei.com> wrote:
>> Hi,Rafael, thanks for you reply.
>> please see my comment.
>>
>> On 2017/9/29 7:13, Rafael J. Wysocki wrote:
>>> On Thu, Sep 28, 2017 at 5:54 AM, Qiang <zhengqiang10@huawei.com> wrote:
>>>> PCIe PME and hot plug share same interrupt number. In some special case,
>>>> Link down event cause hot plug interrupt, devices is not disconnected,
>>>> But read config will return 0xff.
>>>>
>>>> In that case, PME work function will run and not return Because
>>>> Root Status PME bit always 1 and can not be cleared.
>>>>
>>>> This patch add Root Status check in PME interrupt handler,
>>>> Just do same as pciehp isr Slot status check.
>>>>
>>>> Signed-off-by: Qiang Zheng <zhengqiang10@huawei.com>
>>>> ---
>>>>  drivers/pci/pcie/pme.c | 2 +-
>>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>>
>>>> diff --git a/drivers/pci/pcie/pme.c b/drivers/pci/pcie/pme.c
>>>> index fafdb16..67144f5 100644
>>>> --- a/drivers/pci/pcie/pme.c
>>>> +++ b/drivers/pci/pcie/pme.c
>>>> @@ -273,7 +273,7 @@ static irqreturn_t pcie_pme_irq(int irq, void *context)
>>>>         spin_lock_irqsave(&data->lock, flags);
>>>>         pcie_capability_read_dword(port, PCI_EXP_RTSTA, &rtsta);
>>>>
>>>> -       if (!(rtsta & PCI_EXP_RTSTA_PME)) {
>>>> +       if (rtsta == (u32)(~0) || !(rtsta & PCI_EXP_RTSTA_PME)) {
>>>>                 spin_unlock_irqrestore(&data->lock, flags);
>>>>                 return IRQ_NONE;
>>>>         }
>>>
>>> OK, but isn't there a symbol for all ones in u32?  And if there isn't,
>>> what about adding it?
>>>
>> Here is Root Status register definitions.
> 
> Well, that's not what I meant.
> 
> You have "rtsta == (u32)(~0)" in the patch above, but you could do
> "rtsta == U32_MAX" instead.
> 
Sure, I will change it next patch .Thanks.

> Thanks,
> Rafael
> 
> .
>