On Tue, Jan 14, 2025 at 12:15:18PM -0600, Bjorn Helgaas wrote:
> On Tue, Dec 10, 2024 at 04:15:56PM +0800, Richard Zhu wrote:
> > On i.MX8QM, PCIe link can't be re-established again in
> > dw_pcie_resume_noirq(), if the LTSSM_EN bit is not cleared properly in
> > dw_pcie_suspend_noirq().
> >
> > Add dw_pcie_stop_link() into dw_pcie_suspend_noirq() to fix this issue and
> > keep symmetric in suspend/resume function since there is
> > dw_pcie_start_link() in dw_pcie_resume_noirq().
> >
> > Fixes: 4774faf854f5 ("PCI: dwc: Implement generic suspend/resume functionality")
> > Signed-off-by: Richard Zhu <hongxing.zhu@xxxxxxx>
> > Reviewed-by: Manivannan Sadhasivam <manivannan.sadhasivam@xxxxxxxxxx>
> > ---
> > drivers/pci/controller/dwc/pcie-designware-host.c | 1 +
> > 1 file changed, 1 insertion(+)
> >
> > diff --git a/drivers/pci/controller/dwc/pcie-designware-host.c b/drivers/pci/controller/dwc/pcie-designware-host.c
> > index f882b11fd7b94..f56cb7b9e6f99 100644
> > --- a/drivers/pci/controller/dwc/pcie-designware-host.c
> > +++ b/drivers/pci/controller/dwc/pcie-designware-host.c
> > @@ -1001,6 +1001,7 @@ int dw_pcie_suspend_noirq(struct dw_pcie *pci)
> > return ret;
> > }
> >
> > + dw_pcie_stop_link(pci);
>
> We should try to avoid changes to the generic DWC path just to
> accommodate one controller. Since other DWC-based controllers
> apparently don't need dw_pcie_stop_link() here, this seems like it
> might be the wrong place for this change.
>
> If doing dw_pcie_stop_link() here is really helpful for all DWC
> controllers, this would be fine, but the commit log should then explain
> why it helps everybody, not why one particular controller benefits.
It should be for all dwc controllers although find such problem at i.MX8QM
platfrom. It should keep symmetric between suspend/resume function.
So far only layerscape and i.MX platform use these common functions. Other
dwc platform still have not switched to this common function yet.
Frank
>
> If it's only needed for i.MX8QM, we do already have a
> controller-specific hook (.deinit()) just below; maybe that could be
> used?
>
> > if (pci->pp.ops->deinit)
> > pci->pp.ops->deinit(&pci->pp);
> >
> > --
> > 2.37.1
> >
