Re: [PATCH RFC 0/2] Generate device tree node for pci devices

[Sonal Santan]

On 10/6/22 08:10, Rob Herring wrote:
> On Fri, Sep 30, 2022 at 2:29 PM Sonal Santan <sonal.santan@xxxxxxx> wrote:
> > On 9/26/22 15:44, Rob Herring wrote:
> > > On Fri, Sep 16, 2022 at 6:15 PM Frank Rowand <frowand.list@xxxxxxxxx> wrote:
> > > > On 8/29/22 16:43, Lizhi Hou wrote:
> > > > > This patch series introduces OF overlay support for PCI devices which
> > > > > primarily addresses two use cases. First, it provides a data driven method
> > > > > to describe hardware peripherals that are present in a PCI endpoint and
> > > > > hence can be accessed by the PCI host. An example device is Xilinx/AMD
> > > > > Alveo PCIe accelerators. Second, it allows reuse of a OF compatible
> > > > > driver -- often used in SoC platforms -- in a PCI host based system. An
> > > > > example device is Microchip LAN9662 Ethernet Controller.
> > > > >
> > > > > This patch series consolidates previous efforts to define such an
> > > > > infrastructure:
> > > > > https://lore.kernel.org/lkml/20220305052304.726050-1-lizhi.hou@xxxxxxxxxx/
> > > > > https://lore.kernel.org/lkml/20220427094502.456111-1-clement.leger@xxxxxxxxxxx/
> > > > >
> > > > > Normally, the PCI core discovers PCI devices and their BARs using the
> > > > > PCI enumeration process. However, the process does not provide a way to
> > > > > discover the hardware peripherals that are present in a PCI device, and
> > > > > which can be accessed through the PCI BARs. Also, the enumeration process
> > > > > does not provide a way to associate MSI-X vectors of a PCI device with the
> > > > > hardware peripherals that are present in the device. PCI device drivers
> > > > > often use header files to describe the hardware peripherals and their
> > > > > resources as there is no standard data driven way to do so. This patch> series proposes to use flattened device tree blob to describe the
> > > > > peripherals in a data driven way.
> > > >
> > > > > Based on previous discussion, using
> > > > > device tree overlay is the best way to unflatten the blob and populate
> > > > > platform devices.
> > > >
> > > > I still do not agree with this statement.  The device tree overlay
> > > > implementation is very incomplete and should not be used until it
> > > > becomes more complete.  No need to debate this right now, but I don't want
> > > > to let this go unchallenged.
> > >
> > > Then we should remove overlay support. The only way it becomes more
> > > complete is having actual users.
> > >
> > > But really, whether this is the right solution to the problem is
> > > independent of the state of kernel overlay support.
> > >
> > > > If there is no base system device tree on an ACPI based system, then I
> > > > am not convinced that a mixed ACPI / device tree implementation is
> > > > good architecture.
> > >
> > > Most/all of this series is needed for a DT system in which the PCI
> > > devices are not populated in the DT.
> > >
> > > >    I might be more supportive of using a device tree
> > > > description of a PCI device in a detached device tree (not linked to
> > > > the system device tree, but instead freestanding).  Unfortunately the
> > > > device tree functions assume a single system devicetree, with no concept
> > > > of a freestanding tree (eg, if a NULL device tree node is provided to
> > > > a function or macro, it often defaults to the root of the system device
> > > > tree).  I need to go look at whether the flag OF_DETACHED handles this,
> > > > or if it could be leveraged to do so.
> > >
> > > Instead of worrying about a theoretical problem, we should see if
> > > there is an actual problem for a user.
> > >
> > > I'm not so worried about DT functions themselves, but places which
> > > have 'if ACPI ... else (DT) ...' paths.
> >
> > Bringing this thread back into focus. Any thoughts on how to move forward?
>
> Reviewers raise concerns/issues and the submitters work to address
> them or explain why they aren't an issue. The submitter has to push
> things forward. That's how the process works.
We are working on updating the patch set to address the feedback. The 
design is still based on device tree overlay infrastructure.

> As I noted, much of this is needed on a DT system with PCI device not
> described in DT. So you could split out any ACPI system support to
> avoid that concern for example. Enabling others to exercise these
> patches may help too. Perhaps use QEMU to create some imaginary
> device.
To verify this patch set, in addition to a x86_64/ACPI based system, we 
also have an AARCH64/DT QEMU setup where we have attached a physical 
Alveo device. We haven't run into any ACPI or DTO issues so far.

Perhaps we can introduce this feature in a phased manner where we first 
enable DT based platforms and then enable ACPI based platforms?

-Sonal
> Rob