Re: [BUG] PCI/USB4: AMD Phoenix 1022:14ef runtime suspend prevents JHL9480 downstream enumeration

[Mario Limonciello <mario.limonciello@amd.com>]

On 9/25/26 06:50, Mika Westerberg wrote:
> Hi,
> 
> On Fri, Sep 25, 2026 at 11:31:55AM +0000, Freeman Z wrote:
>> Hi Mika, Alex and Mario,
>>
>> I tested v7.3-rc4 with the following Ubuntu SAUCE patch from
>> Chia-Lin Kao (AceLan) at Canonical:
>>
>>      thunderbolt: Fix PCIe device enumeration with delayed rescan
>>
>> Ubuntu SRU posting:
>>      https://lists.ubuntu.com/archives/kernel-team/2026-April/166926.html
>>
>> Launchpad:
>>      https://bugs.launchpad.net/bugs/2139572
>>
>> The Ubuntu posting says this was cherry-picked from the corresponding
>> upstream submission:
>>
>>      https://lore.kernel.org/lkml/20260121060857.237777-1-acelan.kao@canonical.com/T/#u
>>
>> The patch schedules a delayed pci_rescan_bus() 300 ms after the
>> Thunderbolt PCIe tunnel has been activated, to recover devices that
>> pciehp did not enumerate.
> 
> Oh I remember that. It's a horrible hack that is never going to applied.
> 
> Based on this we now know why it "works" in Ubuntu. Thanks for finding it.

Yeah; great finding, this certainly explains the difference.

> 
> I think the real issue is that the PME is not propagated as it should,
> either missing wake or something else but that's what the AMD folks can
> hopefully help us to figure out.

My guess is that there is a race condition of some sort happening.

If we can come up with a repro on our side we can dig further into it.

Alex already tried to downgrade a reference board to match the AGESA 
version in your BIOS, but that wasn't enough.

Can you please share your specific kconfig that you're using?  Maybe 
aligning that can help.