Re: [PATCH 2/2] xhci: drop 120ms U3 LFPS wake retry wait from runtime resume

[Mathias Nyman <mathias.nyman@linux.intel.com>]

On 9/16/26 20:47, Lovekesh Solanki wrote:
> On Fri, Sep 04, 2026 at 06:43:43PM +0530, Lovekesh Solanki wrote:
>> On Thu, Sep 03, 2026 at 10:42:48AM +0300, Mathias Nyman wrote:
>>> On 9/2/26 21:21, Lovekesh Solanki wrote:
>>>> On Wed, Sep 02, 2026 at 03:10:41PM +0300, Mathias Nyman wrote:
>>>>> It's possible PCI side or pm core already provides some easy way for xhci driver to check this.
>>>>> This specific issue is currently not on my todo list, yet.
>>>> Alright, would you like to me to work on this? I could look into
>>>> that side and possibly come up with an approach.
>>>
>>> Yes, please, that would be great.
>> I took a look at the PCI / PM side, and it seems to me there isn't any
>> reliable approach for xhci driver to use right now.
>>
>> The pci_check_pme_status() both checks and clears the pme status bit
>> and the pci core devours it inside pci_pme_wakeup(), and
>> pm_request_resume is async so xhci's runtime_resume may run much later
>> after the bit is already gone. We can't simply export pci_check_pme_status()
>> that wouldn't give any useful info.
>>
>> One solution I can think of would be adding a pme_wake_detected bitfield
>> on pci_dev struct, we set it in pci_pme_wakeup() when PME is confirmed
>> and we clear it in pci_pm_runtime_resume() after the driver has
>> returned.
>> Then, we could directly read via pdev->pme_wake_detected in its own
>> resume.
>>
>> What are your thoughts on this?
> 
> Hey Mathias,
> 
> Just wondering if you have looked into the proposal above.
> I suggested adding a new field to pci_dev struct for knowing whether a
> resume came from PCI PME side.
> 
> Would it be better to send it as an RFC too?

I don't think there is a need for a new PCI bus specific implementation.
pm core then probably handles this already in a generic way.

PCI core probably calls pm core on a PME, which then starts the wakeup,
pm core very likely keeps track of which device triggered the wakeup

The wakeup_source structure under power in the device structure (dev->power.wakeup)
looks promising.
Maybe we can find what we need from there. Hope its fields aren't cleared at the
time xhci resume is called.

Thanks
Mathias