On Fri, Mar 22, 2024 at 12:53:50PM +0100, Niklas Cassel wrote:
> On Thu, Mar 14, 2024 at 01:18:06PM +0530, Manivannan Sadhasivam wrote:
> > "core_init_notifier" flag is set by the glue drivers requiring refclk from
> > the host to complete the DWC core initialization. Also, those drivers will
> > send a notification to the EPF drivers once the initialization is fully
> > completed using the pci_epc_init_notify() API. Only then, the EPF drivers
> > will start functioning.
> >
> > For the rest of the drivers generating refclk locally, EPF drivers will
> > start functioning post binding with them. EPF drivers rely on the
> > 'core_init_notifier' flag to differentiate between the drivers.
> > Unfortunately, this creates two different flows for the EPF drivers.
> >
> > So to avoid that, let's get rid of the "core_init_notifier" flag and follow
> > a single initialization flow for the EPF drivers. This is done by calling
> > the dw_pcie_ep_init_notify() from all glue drivers after the completion of
> > dw_pcie_ep_init_registers() API. This will allow all the glue drivers to
> > send the notification to the EPF drivers once the initialization is fully
> > completed.
> >
> > Only difference here is that, the drivers requiring refclk from host will
> > send the notification once refclk is received, while others will send it
> > during probe time itself.
> >
> > But this also requires the EPC core driver to deliver the notification
> > after EPF driver bind. Because, the glue driver can send the notification
> > before the EPF drivers bind() and in those cases the EPF drivers will miss
> > the event. To accommodate this, EPC core is now caching the state of the
> > EPC initialization in 'init_complete' flag and pci-ep-cfs driver sends the
> > notification to EPF drivers based on that after each EPF driver bind.
> >
> > Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@xxxxxxxxxx>
> > ---
> > drivers/pci/controller/dwc/pci-dra7xx.c | 2 ++
> > drivers/pci/controller/dwc/pci-imx6.c | 2 ++
> > drivers/pci/controller/dwc/pci-keystone.c | 2 ++
> > drivers/pci/controller/dwc/pci-layerscape-ep.c | 2 ++
> > drivers/pci/controller/dwc/pcie-artpec6.c | 2 ++
> > drivers/pci/controller/dwc/pcie-designware-plat.c | 2 ++
> > drivers/pci/controller/dwc/pcie-keembay.c | 2 ++
> > drivers/pci/controller/dwc/pcie-qcom-ep.c | 1 -
> > drivers/pci/controller/dwc/pcie-rcar-gen4.c | 2 ++
> > drivers/pci/controller/dwc/pcie-tegra194.c | 1 -
> > drivers/pci/controller/dwc/pcie-uniphier-ep.c | 2 ++
> > drivers/pci/endpoint/functions/pci-epf-test.c | 18 +++++-------------
> > drivers/pci/endpoint/pci-ep-cfs.c | 9 +++++++++
> > drivers/pci/endpoint/pci-epc-core.c | 22 ++++++++++++++++++++++
> > include/linux/pci-epc.h | 7 ++++---
> > 15 files changed, 58 insertions(+), 18 deletions(-)
>
> FWIW:
> Tested-by: Niklas Cassel <cassel@xxxxxxxxxx>
>
>
>
> However, when looking at this, I was surprised that you never call something
> that will set:
> init_complete = false;
> from e.g. dw_pcie_ep_deinit() or dw_pcie_ep_cleanup().
>
> I saw that you do seem to set
> init_complete = false;
> in your other follow up series that is based on this one.
>
> What will happen if you run with only this series merged (without your
> follow up series), on a platform that used to have .core_init_notifier?
>
> If you do remove and recreate the symlink on a platform with external
> refclk, since you never set init_complete to false, you could trigger
> EPF core_init callback, e.g. pci_epf_test_core_init() to be called,
> which will do DBI writes even when there is no refclk.
>
> E.g. (on a platform with external refclk):
> 1) Create symlink to pci-epf-test in configfs.
> 2) Start RC, your EPC driver will call ep_init_notifiy() when perst
> deasserts.
> 3) Run pci-epf-test.
> 4) Remove the pci-epf-test symlink
> 5) Shutdown RC
> 6) Create symlink to pci-epf-test in configfs.
> This will see that init_complete is true, and will do DBI writes
> which will crash your system, since you don't have a refclk.
>
> Perhaps you should move the patch that calls a function that sets
> init_complete = false;
> to this series, so that this crash is not possible?
>
Good catch! But moving that patch to this series requires moving some other
patches as well. So in the meantime, I'll set this flag to false in
dw_pcie_ep_cleanup().
[...]
> > diff --git a/drivers/pci/endpoint/functions/pci-epf-test.c b/drivers/pci/endpoint/functions/pci-epf-test.c
> > index 18c80002d3bd..fc0282b0d626 100644
> > --- a/drivers/pci/endpoint/functions/pci-epf-test.c
> > +++ b/drivers/pci/endpoint/functions/pci-epf-test.c
[...]
> > - if (!core_init_notifier) {
> > - ret = pci_epf_test_core_init(epf);
> > - if (ret)
> > - return ret;
> > - }
> > -
>
> While you did fix up all DWC based drivers, the non-DWC EPC drivers that
> did not have epc_features->core_init_notifier before this patch:
>
> drivers/pci/controller/cadence/pcie-cadence-ep.c:#include <linux/pci-epc.h>
> drivers/pci/controller/pcie-rcar-ep.c:#include <linux/pci-epc.h>
> drivers/pci/controller/pcie-rockchip-ep.c:#include <linux/pci-epc.h>
>
> I don't think that they will work with pci-epf-test anymore, since AFAICT,
> you did not add a call to: pci_epc_init_notify() or similar in these EPC drivers.
> (Like this patch does to all the DWC-based drivers without a core_init_notifier.)
>
Doh, yeah I completely missed these. Thanks for pointing out. Will add the
notify_init call in next version.
- Mani
--
மணிவண்ணன் சதாசிவம்
