Re: [PATCH v2] driver core: platform: Setup device MSI domain just before driver probe

On Mon, Jul 27, 2026 at 5:46 PM Robin Murphy <[email protected]> wrote:
>
> On 27/07/2026 9:40 am, Anup Patel wrote:
> > On RISC-V, the MSI controller (aka RISC-V IMSIC) is probed as a regular
> > platform device and MSI client drivers are always probed after the MSI
> > controller driver using fw_devlink. Unfortunately, this is not sufficient
> > to ensure device MSI domain is set for MSI client devices before driver
> > probe because OF framework sets device MSI domain at the time of platform
> > device creation whereas ACPI framework expects arch specific code to set
> > the device MSI domain at the time of platform device creation.
> >
> > Currently, to address the problem of device MSI domain being not set
> > correctly, various RISC-V MSI client drivers explicitly set device
> > MSI domain in the driver probe function using below code pattern:
> >
> >       /*
> >        * The device MSI domain for platform devices on RISC-V architecture
> >        * is only available after the MSI controller driver is probed so,
> >        * explicitly configure here.
> >        */
> >       if (!dev_get_msi_domain(dev)) {
> >               /*
> >                * The device MSI domain for OF devices is only set at the
> >                * time of populating/creating OF device. If the device MSI
> >                * domain is discovered later after the OF device is created
> >                * then we need to set it explicitly before using any platform
> >                * MSI functions.
> >                */
> >               if (is_of_node(fwnode)) {
> >                       of_msi_configure(dev, dev_of_node(dev));
> >               } else if (is_acpi_device_node(fwnode)) {
> >                       struct irq_domain *msi_domain;
> >                       msi_domain = irq_find_matching_fwnode(imsic_acpi_get_fwnode(dev),
> >                                                             DOMAIN_BUS_PLATFORM_MSI);
> >                       dev_set_msi_domain(dev, msi_domain);
> >               }
> >
> >               if (!dev_get_msi_domain(dev))
> >                       return -EPROBE_DEFER;
>
> I stand by my previous comment that relying on ACPI initcall order and
> assuming fw_devlink is enabled doesn't really seem robust enough to
> justify dropping proper handling of this case, especially if pretending
> it's a generic interface.

I think you still fail to understand the root-cause of the msi_domain not
being set correctly in the struct device. There is no issue with the probe
order (or probe deferral) rather the issue is msi_domain being set very
early by OF / ACPI framework at the time of device creation when there
is no MSI domain registered.

Let's take the case of OF based platform devices with RISC-V:

1) The OF framework set the msi_domain at the time of platform
    device creation but unfortunately the RISC-V MSI controller
    driver is also a platform device (aka RISC-V IMSIC driver) which
    is probed much latter hence the msi_domain for all OF platform
    devices is set to NULL.
    (Refer call path,  of_platform_device_create_pdata() ->
of_msi_configure() -> dev_set_msi_domain()

2) The fw_devlink based probe ordering in the Linux DD framework
    ensures that RISC-V MSI controller driver is probed before the
    MSI client drivers (such as RISC-V APLIC MSI driver, RISC-V
    system MSI driver, etc) but still when these MSI client drivers
    are probed the msi_domain remains NULL for the platform device
    of MSI client.

3) Due to #1 and #2 above, the MSI client driver have to explicitlly
    call of_msi_configure() on RISC-V because no one updates the
    msi_domain of all platform devices after RISC-V MSI controller
    registers a MSI irqdomain.

This patch eliminates the #3 above by calling of_msi_configure()
right before probing the OF platform device.

Regards,
Anup

_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv