[PATCH] pinctrl: denverton: Fix probe of MFD-created platform device

[PATCH] pinctrl: denverton: Fix probe of MFD-created platform device

[Radek Válko]

The Denverton pinctrl device can be instantiated by the Intel LPC
driver as an MFD platform device named "denverton-pinctrl".

On affected systems the platform device does not carry the INTC3000
ACPI match data itself. As a result, intel_pinctrl_probe_by_hid()
fails to obtain the SoC data using device_get_match_data() and
returns -ENODATA.

This results in:

  denverton-pinctrl denverton-pinctrl.0:
  probe with driver denverton-pinctrl failed with error -61

Denverton has a single fixed intel_pinctrl_soc_data instance. Pass
dnv_soc_data directly to intel_pinctrl_probe() instead of requiring
firmware match data for the MFD-created platform device.

Tested on an Intel Atom C3000 system with LPC/eSPI device 8086:19dc.
Before this change the Denverton pinctrl driver failed to probe and
no Denverton GPIO controller was registered. After the change,
denverton-pinctrl.0 registers with pinmux and pinconf support and
exposes 154 GPIO lines.

Signed-off-by: Radek Válko <rvalko@lipicko.cz>
---
 drivers/pinctrl/intel/pinctrl-denverton.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/drivers/pinctrl/intel/pinctrl-denverton.c b/drivers/pinctrl/intel/pinctrl-denverton.c
--- a/drivers/pinctrl/intel/pinctrl-denverton.c
+++ b/drivers/pinctrl/intel/pinctrl-denverton.c
@@ -250,6 +250,11 @@ static const struct intel_pinctrl_soc_data dnv_soc_data = {
 	.ncommunities = ARRAY_SIZE(dnv_communities),
 };
 
+static int dnv_pinctrl_probe(struct platform_device *pdev)
+{
+	return intel_pinctrl_probe(pdev, &dnv_soc_data);
+}
+
 static const struct acpi_device_id dnv_pinctrl_acpi_match[] = {
 	{ "INTC3000", (kernel_ulong_t)&dnv_soc_data },
 	{ }
@@ -263,7 +268,7 @@ static const struct platform_device_id dnv_pinctrl_platform_ids[] = {
 MODULE_DEVICE_TABLE(platform, dnv_pinctrl_platform_ids);
 
 static struct platform_driver dnv_pinctrl_driver = {
-	.probe = intel_pinctrl_probe_by_hid,
+	.probe = dnv_pinctrl_probe,
 	.driver = {
 		.name = "denverton-pinctrl",
 		.acpi_match_table = dnv_pinctrl_acpi_match,
-- 
2.47.0

Re: [PATCH] pinctrl: denverton: Fix probe of MFD-created platform device

[Andy Shevchenko]

On Thu, Aug 27, 2026 at 08:42:58PM +0200, Radek Válko wrote:
> The Denverton pinctrl device can be instantiated by the Intel LPC
> driver as an MFD platform device named "denverton-pinctrl".
> 
> On affected systems the platform device does not carry the INTC3000
> ACPI match data itself. As a result, intel_pinctrl_probe_by_hid()
> fails to obtain the SoC data using device_get_match_data() and
> returns -ENODATA.
> 
> This results in:
> 
>   denverton-pinctrl denverton-pinctrl.0:
>   probe with driver denverton-pinctrl failed with error -61
> 
> Denverton has a single fixed intel_pinctrl_soc_data instance. Pass
> dnv_soc_data directly to intel_pinctrl_probe() instead of requiring
> firmware match data for the MFD-created platform device.
> 
> Tested on an Intel Atom C3000 system with LPC/eSPI device 8086:19dc.
> Before this change the Denverton pinctrl driver failed to probe and
> no Denverton GPIO controller was registered. After the change,
> denverton-pinctrl.0 registers with pinmux and pinconf support and
> exposes 154 GPIO lines.

Thanks for the report.

Can you instead try to replace device_get_match_data() in
intel_pinctrl_probe_by_hid() by intel_pinctrl_get_soc_data() in the way how
it's done in intel_pinctrl_probe_by_uid()?

-- 
With Best Regards,
Andy Shevchenko

Re: [PATCH] pinctrl: denverton: Fix probe of MFD-created platform device

[Andy Shevchenko]

On Thu, Aug 27, 2026 at 09:52:40PM +0300, Andy Shevchenko wrote:
> On Thu, Aug 27, 2026 at 08:42:58PM +0200, Radek Válko wrote:
> > The Denverton pinctrl device can be instantiated by the Intel LPC
> > driver as an MFD platform device named "denverton-pinctrl".
> > 
> > On affected systems the platform device does not carry the INTC3000
> > ACPI match data itself. As a result, intel_pinctrl_probe_by_hid()
> > fails to obtain the SoC data using device_get_match_data() and
> > returns -ENODATA.
> > 
> > This results in:
> > 
> >   denverton-pinctrl denverton-pinctrl.0:
> >   probe with driver denverton-pinctrl failed with error -61
> > 
> > Denverton has a single fixed intel_pinctrl_soc_data instance. Pass
> > dnv_soc_data directly to intel_pinctrl_probe() instead of requiring
> > firmware match data for the MFD-created platform device.
> > 
> > Tested on an Intel Atom C3000 system with LPC/eSPI device 8086:19dc.
> > Before this change the Denverton pinctrl driver failed to probe and
> > no Denverton GPIO controller was registered. After the change,
> > denverton-pinctrl.0 registers with pinmux and pinconf support and
> > exposes 154 GPIO lines.
> 
> Thanks for the report.
> 
> Can you instead try to replace device_get_match_data() in
> intel_pinctrl_probe_by_hid() by intel_pinctrl_get_soc_data() in the way how
> it's done in intel_pinctrl_probe_by_uid()?

Ah, it won't work as direct replacement. So, the intel_pinctrl_probe_by_hid()
should be modified accordingly. I will send a formal patch tomorrow.

-- 
With Best Regards,
Andy Shevchenko