On 15 October 2025 at 03:51 pm, Herve Codina [off-list ref] wrote:

Hi Christian,

On Wed, 15 Oct 2025 15:14:28 +0200
Christian Zigotzky [off-list ref] wrote:

quoted

quoted

Am 15 October 2025 at 02:59 pm, Herve Codina [off-list ref] wrote:

Also when it boots, it is not easy to know about the problem root cause.

In my case, it was not obvious to make the relationship on my side between
my ping timings behavior and ASPM.

Of course 'git bisect' helped a lot but can we rely on that for the average
user?

Best regards,
Hervé  

I think I will revert these modifications for the RC2.

I don't know what is the future of those modifications but maybe instead
of fully reverting them, maybe you could perform calls to
- pcie_clkpm_override_default_link_state() and
- pcie_aspm_override_default_link_state()
only if a new Kconfig symbol is enabled.

Of course this symbols will be disabled by default but if you want some
people to test behavior, it could be interesting to have the code
available in the kernel.

I don't know, this was just an idea.

Of course, reverting the patch is simpler than adding this new Kconfig
symbol.

Best regards,
Hervé

Hi Hervé,

Do you mean an option in the kernel config? If yes, then it is a great idea.

@Mani
Could you please create an option in the kernel config that enable or disable the power management for PCI and PCI Express?
If yes, then I don’t need to revert the changes due to boot issues and less performance.

Just for info: Simon Richter wrote:

Intel B580 on TalosII has trouble growing the BAR with 6.18.

Power management changes should not affect PPC because it is broken anyway — PCIe bridges are assumed to be managed by OpenFirmware, so the "pcieport" driver is not registered, so bridges have no power management, so downstream devices cannot enable it either.

- - -

Thanks,
Christian