Re: [PATCH V1] PCI: designware-ep: Fix DBI access before core init

[Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>]

On Mon, Aug 01, 2022 at 02:27:14PM -0600, Rob Herring wrote:
> On Sat, Jul 30, 2022 at 8:50 AM Manivannan Sadhasivam
> <manivannan.sadhasivam@linaro.org> wrote:
> >
> > On Fri, Jul 29, 2022 at 05:44:04PM -0500, Bjorn Helgaas wrote:
> > > [+cc Xiaowei (author of 6bfc9c3a2c70), Hou (author of 8bcca2658558)]
> > >
> > > On Thu, Jul 28, 2022 at 05:56:28PM +0530, Vidya Sagar wrote:
> > > > On 7/28/2022 3:44 AM, Bjorn Helgaas wrote:
> > > > > On Wed, Jun 22, 2022 at 09:31:33AM +0530, Vidya Sagar wrote:
> > > > > > Platforms that cannot support their core initialization without the
> > > > > > reference clock from the host, implement the feature 'core_init_notifier'
> > > > > > to indicate the DesignWare sub-system about when their core is getting
> > > > > > initialized. Any accesses to the core (Ex:- DBI) would result in system
> > > > > > hang in such systems (Ex:- tegra194). This patch moves any access to the
> > > > > > core to dw_pcie_ep_init_complete() API which is effectively called only
> > > > > > after the core initialization.
> 
> > >   6) What's going on with the CORE_INIT and LINK_UP notifiers?
> > >      dw_pcie_ep_init_notify() is only called by qcom and tegra.
> > >      dw_pcie_ep_linkup() is only called by dra7xx, qcom, and tegra.
> > >      As far as I can tell, nobody at all registers to handle those
> > >      events except a test.  I think it's pointless to have that code
> > >      if nobody uses it.
> > >
> >
> > I have submitted an actual driver that makes use of these notifiers:
> > https://lore.kernel.org/lkml/20220502060611.58987-9-manivannan.sadhasivam@linaro.org/
> 
> Notifiers aren't the best interface in the kernel. I think they are
> best used if there's no real linkage between the sender and receiver.
> For an EPC and EPF that's a fixed interface, so define a proper
> interface.
> 

Fair point! The use of notifiers also suffer from an issue where the notifier
chain in EPC is atomic but the EPF calls some of the functions like
pci_epc_write_header() could potentially sleep.

I'll try to come up with an interface.

Thanks,
Mani

> Rob

-- 
மணிவண்ணன் சதாசிவம்