Re: [REGRESSION] Intel Wireless adapter is not detected until suspending to RAM and resuming

["Mario Limonciello (AMD) (kernel.org)" <superm1@kernel.org>]

On 10/20/2025 7:56 AM, Adrià Vilanova Martínez wrote:
> On Sun, Oct 19, 2025 at 07:25:08PM -0500, Mario Limonciello wrote:
>> Thanks, knowing that pcie_aspm=off helps I think we should compare output
>> for:
>>
>> # sudo lspci -vvnn
> 
> Sure, I'm attaching the outputs of this command for all the scenarios.
> There are some differences, so it seems promising.
> 

Surprisingly there is nothing different about ASPM though.  It's all 
PCI-PM differences.

Looking at your log again I noticed this from the bridge:

pcieport 0000:00:1c.0: pciehp: Slot(0): Card not present
pcieport 0000:00:1c.0: pciehp: Slot(0): Card present
pcieport 0000:00:1c.0: pciehp: Slot(0): Link Up
...
pcieport 0000:00:1c.0: pciehp: Slot(0): No device found
...
(suspend)
...
pcieport 0000:00:1c.0: pciehp: Slot(0): Card present
pcieport 0000:00:1c.0: pciehp: Slot(0): Link Up


> I'm building the Kernel on the following commits:
> 
> - "Kernel without 4d4c10f763 and 907a7a2e5b": 1c64efcb08, applying on
> top reverts for these 2 commits. [locally compiled version
> 6.18.0-rc1-local-reverted-pci-issues-00351-gbbaff7ff47dd]
> - "Kernel with 4d4c10f763 and 907a7a2e5b": 1c64efcb08 (last commit I
> pulled from mainline last week). [locally compiled version ???]
> 
>> In the following cases (all without pcie_aspm=off):
>>
>> 1) At bootup; a kernel without 4d4c10f763 and 907a7a2e5b
> 
> See 01_lspci_bootup_without_4d4c10f763_907a7a2e5b.txt
> 

OK so the bridge at 00:1c.0:
L1SubCtl1: PCI-PM_L1.2+ PCI-PM_L1.1+ ASPM_L1.2- ASPM_L1.1-

01:00.0 is present

>> 2) At bootup; a kernel with 4d4c10f763 and 907a7a2e5b
> 
> See 02_lspci_bootup_with_4d4c10f763_907a7a2e5b.txt
> 

OK so the bridge at 00:1c.0:
L1SubCtl1: PCI-PM_L1.2- PCI-PM_L1.1- ASPM_L1.2- ASPM_L1.1-

01:00.0 is NOT present

>> 3) After suspend/resume; a kernel without 4d4c10f763 and 907a7a2e5b4
> 
> See 03_lspci_after_suspend_resume_without_4d4c10f763_907a7a2e5b.txt
> 

OK so the bridge at 00:1c.0:
L1SubCtl1: PCI-PM_L1.2+ PCI-PM_L1.1+ ASPM_L1.2- ASPM_L1.1-

01:00.0 is present

>> 4) After suspend/resume; a kernel with 4d4c10f763 and 907a7a2e5b
> 
> See 04_lspci_after_suspend_resume_with_4d4c10f763_907a7a2e5b.txt
> 

OK so the bridge at 00:1c.0:
L1SubCtl1: PCI-PM_L1.2- PCI-PM_L1.1- ASPM_L1.2- ASPM_L1.1-

01:00.0 is present

> Again, thank you so much! I really appreciate your help in
> troubleshooting this.

My interpretation is that the BIOS by default starts with PCI PM 
enabled.  When you test without 4d4c10f763 and 907a7a2e5b it will stay 
enabled.  But when those commits are present it gets disabled when going 
to D0 and that causes device to drop off the bus.

How about with pcie_port_pm=off instead of pcie_aspm=off?  Do things work?

My current thought is that the change (setting to D0 explicitly at 
boot-up) exposed a bug in the platform.  But the fact that it works 
without ASPM is confusing to me.

Bjorn - any thoughts here?

> 
> PS: I'm trimming the email quotes as per
> https://subspace.kernel.org/etiquette.html#trim-your-quotes-when-replying.
> I've never done this before and it feels wrong, but it is indeed easier
> to follow the conversation if I do this.