On Wed, Jul 16, 2025 at 12:23:54PM GMT, Krishna Chaitanya Chundru wrote:
>
>
> On 7/16/2025 12:16 PM, Manivannan Sadhasivam wrote:
> > On Wed, Jul 16, 2025 at 10:24:23AM GMT, Krishna Chaitanya Chundru wrote:
> > >
> > >
> > > On 7/15/2025 4:06 PM, Manivannan Sadhasivam wrote:
> > > > On Tue, Jul 15, 2025 at 11:54:48AM GMT, Konrad Dybcio wrote:
> > > > > On 7/14/25 8:01 PM, Manivannan Sadhasivam wrote:
> > > > > > It allows us to group all the settings that need to be done when a PCI
> > > > > > device is attached to the bus in a single place.
> > > > > >
> > > > > > Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@xxxxxxxxxxxxxxxx>
> > > > > > ---
> > > > > > drivers/pci/controller/dwc/pcie-qcom.c | 3 +--
> > > > > > 1 file changed, 1 insertion(+), 2 deletions(-)
> > > > > >
> > > > > > diff --git a/drivers/pci/controller/dwc/pcie-qcom.c b/drivers/pci/controller/dwc/pcie-qcom.c
> > > > > > index b4993642ed90915299e825e47d282b8175a78346..b364977d78a2c659f65f0f12ce4274601d20eaa6 100644
> > > > > > --- a/drivers/pci/controller/dwc/pcie-qcom.c
> > > > > > +++ b/drivers/pci/controller/dwc/pcie-qcom.c
> > > > > > @@ -1616,8 +1616,6 @@ static irqreturn_t qcom_pcie_global_irq_thread(int irq, void *data)
> > > > > > pci_lock_rescan_remove();
> > > > > > pci_rescan_bus(pp->bridge->bus);
> > > > > > pci_unlock_rescan_remove();
> > > > > > -
> > > > > > - qcom_pcie_icc_opp_update(pcie);
> > > > > > } else {
> > > > > > dev_WARN_ONCE(dev, 1, "Received unknown event. INT_STATUS: 0x%08x\n",
> > > > > > status);
> > > > > > @@ -1765,6 +1763,7 @@ static int pcie_qcom_notify(struct notifier_block *nb, unsigned long action,
> > > > > > switch (action) {
> > > > > > case BUS_NOTIFY_BIND_DRIVER:
> > > > > > qcom_pcie_enable_aspm(pdev);
> > > > > > + qcom_pcie_icc_opp_update(pcie);
> > > > >
> > > > > So I assume that we're not exactly going to do much with the device if
> > > > > there isn't a driver for it, but I have concerns that since the link
> > > > > would already be established(?), the icc vote may be too low, especially
> > > > > if the user uses something funky like UIO
> > > > >
> > > >
> > > > Hmm, that's a good point. Not enabling ASPM wouldn't have much consequence, but
> > > > not updating OPP would be.
> > > >
> > > > Let me think of other ways to call these two APIs during the device addition. If
> > > > there are no sane ways, I'll drop *this* patch.
> > > >
> > > How about using enable_device in host bridge, without pci_enable_device
> > > call the endpoints can't start the transfers. May be we can use that.
> > >
> >
> > Q: Who is going to call pci_enable_device()?
> > A: The PCI client driver
> >
> > This is same as relying on BUS_NOTIFY_BIND_DRIVER notifier.
> >
> userspace can enable device using sysfs[1] without attaching
> any kernel drivers.
>
But that's not a common usecase. Even so, we cannot insist users to write to the
sysfs knob to let ASPM/OPP work without a driver.
- Mani
--
மணிவண்ணன் சதாசிவம்
