Re: [PATCH] mfd: Add Simple PCI MFD driver

[Rob Herring <robh@kernel.org>]

On Mon, Jan 23, 2023 at 8:32 AM Vincent Whitchurch
<vincent.whitchurch@axis.com> wrote:
>
> Add a PCI driver which registers all child nodes specified in the
> devicetree.  It will allow platform devices to be used on virtual
> systems which already support PCI and devicetree, such as UML with
> virt-pci.

There's similar work underway for Xilinx/AMD PCIe FPGAs[1]. It's the
same thing really. Non-discoverable things downstream of a PCI device.
There's also a desire for that to work on non-DT (ACPI) based hosts.
While UML supports DT, that's currently only for the unittest AFAIK.
So it's more like a non-DT host. How does the DT get populated for UML
for this to work?

Can you provide details on the actual h/w you want to use. What
problem are you trying to solve?

Really, what I want to see here is everyone interested in this feature
to work together on it. Not just creating a one-off solution for their
1 use case that's a subset of a bigger solution.

> The driver has no id_table by default; user space needs to provide one
> using the new_id mechanism in sysfs.

But your DT will have the id in it already. Wouldn't you rather
everything work without userspace intervention? I can't imagine the
list here would be too long.

>
> Signed-off-by: Vincent Whitchurch <vincent.whitchurch@axis.com>
> ---
>  drivers/mfd/Kconfig          | 11 +++++++++++
>  drivers/mfd/Makefile         |  1 +
>  drivers/mfd/simple-mfd-pci.c | 21 +++++++++++++++++++++
>  3 files changed, 33 insertions(+)
>
> diff --git a/drivers/mfd/Kconfig b/drivers/mfd/Kconfig
> index 30db49f31866..1e325334e9ae 100644
> --- a/drivers/mfd/Kconfig
> +++ b/drivers/mfd/Kconfig
> @@ -1277,6 +1277,17 @@ config MFD_SIMPLE_MFD_I2C
>           sub-devices represented by child nodes in Device Tree will be
>           subsequently registered.
>
> +config MFD_SIMPLE_MFD_PCI
> +       tristate "Simple Multi-Functional Device support (PCI)"
> +       depends on PCI
> +       depends on OF || COMPILE_TEST
> +       help
> +         This enables support for a PCI driver for which any sub-devices
> +         represented by child nodes in the devicetree will be registered.
> +
> +         The driver does not bind to any devices by default; that should
> +         be done via sysfs using new_id.
> +
>  config MFD_SL28CPLD
>         tristate "Kontron sl28cpld Board Management Controller"
>         depends on I2C
> diff --git a/drivers/mfd/Makefile b/drivers/mfd/Makefile
> index 457471478a93..7ae329039a13 100644
> --- a/drivers/mfd/Makefile
> +++ b/drivers/mfd/Makefile
> @@ -268,6 +268,7 @@ obj-$(CONFIG_MFD_QCOM_PM8008)       += qcom-pm8008.o
>
>  obj-$(CONFIG_SGI_MFD_IOC3)     += ioc3.o
>  obj-$(CONFIG_MFD_SIMPLE_MFD_I2C)       += simple-mfd-i2c.o
> +obj-$(CONFIG_MFD_SIMPLE_MFD_PCI)       += simple-mfd-pci.o
>  obj-$(CONFIG_MFD_SMPRO)                += smpro-core.o
>  obj-$(CONFIG_MFD_INTEL_M10_BMC)   += intel-m10-bmc.o
>
> diff --git a/drivers/mfd/simple-mfd-pci.c b/drivers/mfd/simple-mfd-pci.c
> new file mode 100644
> index 000000000000..c5b2540e924a
> --- /dev/null
> +++ b/drivers/mfd/simple-mfd-pci.c
> @@ -0,0 +1,21 @@
> +// SPDX-License-Identifier: GPL-2.0-only
> +
> +#include <linux/module.h>
> +#include <linux/of_platform.h>
> +#include <linux/pci.h>
> +
> +static int simple_mfd_pci_probe(struct pci_dev *pdev,
> +                               const struct pci_device_id *id)
> +{
> +       return devm_of_platform_populate(&pdev->dev);

Really, this could be anything in the child DT. Not just what Linux
classifies as an MFD. So maybe drivers/mfd is not the right place.

Rob

[1] https://lore.kernel.org/all/1674183732-5157-1-git-send-email-lizhi.hou@amd.com/