Re: [PATCH v2 1/2] PCI/ASPM: Override the ASPM and Clock PM states set by BIOS for devicetree platforms

[Manivannan Sadhasivam]

On Thu, Feb 26, 2026 at 04:55:34PM +0000, Jon Hunter wrote:
> 
> On 26/02/2026 11:08, Manivannan Sadhasivam wrote:
> 
> ...
> 
> > Since we know that ASPM is the issue on your platform and the failure also
> > confirms that ASPM was never enabled before, I'd suggest disabling ASPM for the
> > Root Port as a workaround:
> > 
> > ```
> > diff --git a/drivers/pci/controller/dwc/pcie-tegra194.c b/drivers/pci/controller/dwc/pcie-tegra194.c
> > index 06571d806ab3..f504b4ffbcb6 100644
> > --- a/drivers/pci/controller/dwc/pcie-tegra194.c
> > +++ b/drivers/pci/controller/dwc/pcie-tegra194.c
> > @@ -2499,6 +2499,13 @@ module_platform_driver(tegra_pcie_dw_driver);
> >   MODULE_DEVICE_TABLE(of, tegra_pcie_dw_of_match);
> > +static void tegra_pcie_quirk_disable_aspm(struct pci_dev *dev)
> > +{
> > +       pcie_aspm_remove_cap(dev, PCI_EXP_LNKCAP_ASPM_L1 |
> > +                                 PCI_EXP_LNKCAP_ASPM_L0S);
> > +}
> > +DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_NVIDIA, PCI_ANY_ID, tegra_pcie_quirk_disable_aspm);
> > +
> >   MODULE_AUTHOR("Vidya Sagar <vidyas@xxxxxxxxxx>");
> >   MODULE_DESCRIPTION("NVIDIA PCIe host controller driver");
> >   MODULE_LICENSE("GPL v2");
> > ```
> > 
> > You can use specific Root Port IDs or PCI_ANY_ID depending on the impact. We can
> > also work on fixing the actual issue parallelly.
> 
> Thanks. By default we are building the PCIe driver for Tegra as a module and
> so I am not sure we can use DECLARE_PCI_FIXUP_EARLY() right?
> 

Ah, yes. We cannot use any of these DECLARE_PCI_FIXUP*() in a module anyway :/

> I was just thinking that in pcie_aspm_override_default_link_state() we just
> need a callback to specify the default ASPM override state?
> 

That looks like a dirty hack. Moreover, your platform works perfectly fine with
ASPM. So you should not be turning it off due to one driver behaving
erratically.

As I mentioned in [1], you should try to advertise the
PM_SUSPEND_FLAG_FW_SUSPEND flag for the platform as per [2].

FYI, this is a known issue that is plauging us for so long.
PM_SUSPEND_FLAG_FW_SUSPEND is the cleanest solution we came up with. But
unfortunately, we cannot use it across all of our Qcom SoCs due to firmware not
advertising S2RAM in some of them.

- Mani

[1] https://lore.kernel.org/linux-pci/kkly3z4durpagtenadvmzdpojlctachgfgi2fdapt6zthdl2gx@n2qhmlud2zb7/
[2] https://lore.kernel.org/all/20251231162126.7728-1-manivannan.sadhasivam@xxxxxxxxxxxxxxxx/

-- 
மணிவண்ணன் சதாசிவம்