Re: [PATCH 1/5] platform/x86: intel_lpss: add resource conflict quirk for Dell Latitude 5285

[Andy Shevchenko]

On Thu, Mar 19, 2026 at 05:09:29PM -0700, Thierry Chatard wrote:
> The Dell Latitude 5285 2-in-1 has a BIOS bug where the ACPI GEXP device
> and the I2C4 controller (INT3446) both claim the same MMIO region via the
> shared SB04 variable. This causes intel_lpss_acpi to fail binding to I2C4
> with -EBUSY, preventing the front camera (OV5670) sensor from being
> registered.

> Add a DMI quirk that selects IGNORE_RESOURCE_CONFLICTS for INT3446 on this
> machine, matching the existing pattern used by other LPSS quirks.

Was there any new version? In any case keep me in Cc list for at lease Intel
LPSS parts.

...

>  #include <linux/ioport.h>
>  #include <linux/mod_devicetable.h>
>  #include <linux/module.h>
> +#include <linux/acpi.h>
> +#include <linux/dmi.h>

Keep that in order.

>  #include <linux/pm.h>
>  #include <linux/pm_runtime.h>
>  #include <linux/platform_device.h>

...

> +/* Same as spt_i2c_info but with QUIRK_IGNORE_RESOURCE_CONFLICTS for Dell 5285
> + * where ACPI GEXP device conflicts with I2C4 (INT3446) MMIO resources.
> + */

/*
 * Multi-line comments have to follow the
 * style as in this example.
 */

> +static const struct intel_lpss_platform_info spt_i2c_info_ignore_conflicts = {
> +	.clk_rate = 120000000,
> +	.swnode = &spt_i2c_node,
> +	.quirks = QUIRK_IGNORE_RESOURCE_CONFLICTS,
> +};

No, please just add a quirk into .driver_data of DMI below.

...

> +	/* Apply IGNORE_RESOURCE_CONFLICTS for I2C4 on Dell Latitude 5285.
> +	 * The ACPI GEXP device conflicts with I2C4 (INT3446) MMIO resources
> +	 * due to a BIOS bug where both use the same SB04 variable.
> +	 */

See about multi-line comment style above.

> +	if (data == &spt_i2c_info &&
> +	    acpi_dev_hid_uid_match(ACPI_COMPANION(&pdev->dev), "INT3446", NULL) &&
> +	    dmi_check_system(dell5285_lpss_dmi)) {
> +		dev_info(&pdev->dev, "Dell 5285: applying IGNORE_RESOURCE_CONFLICTS for I2C4\n");
> +		data = &spt_i2c_info_ignore_conflicts;
> +	}

No, just take a quirk and apply it after we dup the info,

>  	info = devm_kmemdup(&pdev->dev, data, sizeof(*info), GFP_KERNEL);
>  	if (!info)
>  		return -ENOMEM;

Somewhere here against 'info'. See how PCI counterpart does.

And better is to create ACPI ID table for that (as we do for PCI case) and use
it as a matching material. So, something like

	if (DMI matches) {
		struct acpi_device_id *id; // maybe even const, don't remember

		id = match_acpi_ID();
		if (id)
			info->quirks = assign the quirk.
	}

-- 
With Best Regards,
Andy Shevchenko