Re: [PATCH] PCI: pciehp: Disable link and turn off slot power while removing

[Mario Limonciello <superm1@kernel.org>]

On 5/7/26 00:03, Mario Limonciello wrote:
> 
> 
> On 5/6/26 23:26, Lukas Wunner wrote:
>> On Wed, May 06, 2026 at 02:16:10PM -0500, Mario Limonciello wrote:
>>> On 5/6/26 12:55, Lukas Wunner wrote:
>>>> Moreover there are products where Presence Detect is hardwired to zero,
>>>> see commit 80696f991424 ("PCI: pciehp: Tolerate Presence Detect 
>>>> hardwired
>>>> to zero").  So we have to rely on a link change to detect that a new 
>>>> device
>>>> has been plugged in after slot poweroff.  That won't work if the 
>>>> link is
>>>> disabled.
>>>
>>> How would you feel about turning off link detect somewhere specific 
>>> to the
>>> shutdown path?
>>
>> I *think* shutdown is performed bottom-up in the hierarchy and so
>> this might work.  But it's not great to introduce such a change
>> for everyone just because a single product needs it. 
> 
> Generally speaking I'm not a fan of Linux changing hardware state and 
> not cleaning up after itself at shutdown.  This can lead to weird 
> behaviors for power consumption, wakeups, or inconsistent init across a 
> kexec vs warm reboot vs cold reboot (that BIOS had a chance to swing 
> it's hammer).
> 
>> It would
>> be better to root-cause the issue. 
> 
> I'll see if we can figure out what's happening with traces on a USB 
> analyzer.  But FWIW it's a Linux only issue :/
> 
>> Maybe the vendor is able to
>> provide a BIOS update which resolves it?
>>
> 
> Given it's Linux only issue; unlikely.

I'm happy to share that we figured out why it was only happening on 
Linux by comparing the traces.  The USB4 CM on Windows issues issues a 
downstream port reset during the tear down process.  This is missing 
from Linux.

This patch helps it.

https://lore.kernel.org/linux-usb/20260603180146.3949001-1-Basavaraj.Natikar@amd.com/