Re: [PATCH v3 2/2] PCI: Fix runtime PM usage count underflow on device unplug

[Lukas Wunner <lukas@wunner.de>]

On Sun, Jun 22, 2025 at 01:39:26PM -0500, Mario Limonciello wrote:
> > > On 6/21/25 2:05 PM, Lukas Wunner wrote:
> > > > So the refcount decrement happens in pcie_portdrv_probe() and
> > > > the refcount increment happens in pcie_portdrv_remove().
> > > > Both times it's conditional on pci_bridge_d3_possible().
> > > > Does that return a different value on probe versus remove?
> 
> I did this check and yes specifically on this PCIe port with the underflow
> the d3 possible lookup returns false during pcie_portdrv_remove().  It
> returns true during pcie_portdrv_probe().

That's not supposed to happen.  The expectation is that
pci_bridge_d3_possible() always returns the same value.

For this reason the return value on ->probe() isn't cached.

So which of the conditions changes in pci_bridge_d3_possible()
on probe versus remove?  Could you instrument each with a printk()
so that we can understand what's going wrong there?

Thanks,

Lukas