Re: [PATCH v3] usb: typec: ucsi: do not register UCSI on Lenovo Legion Pro 7 16IAX10H

[Heikki Krogerus <heikki.krogerus@linux.intel.com>]

On Wed, Sep 02, 2026 at 05:05:31PM +0800, Huang Wei wrote:
> The EC firmware of the Lenovo Legion Pro 7 16IAX10H (DMI product name
> 83F5) has a broken UCSI implementation: GET_PDOS returns no source
> capabilities, GET_CONNECTOR_STATUS reports an incomplete RDO, and any
> UCSI traffic disturbs the EC's autonomous USB-C Power Delivery
> negotiation. With ucsi_acpi bound, chargers connect only briefly and
> then drop, and voltage_now reads 0; with the driver blacklisted, the
> EC negotiates full PD contracts on its own and charging works normally.
> 
> The reporter confirmed the issue on BIOS versions Q7CN44WW and
> Q7CN78WW (2026-04); it is unaffected by the BIOS update between them.
> The AMD sibling Legion Pro 7 16AFR10H (product name 83RU) reportedly
> uses the same EC firmware, but it is left out of the quirk until the
> failure is confirmed on that model.
> 
> Skip probe on machines carrying the known-broken BIOS versions, which
> is equivalent to the blacklist workaround the reporter verified to
> restore charging. Matching the specific BIOS versions rather than the
> whole platform ensures that a future firmware that fixes UCSI is not
> wrongly skipped.
> 
> Reported-by: Alex Bartz <alex@alstergee.com>
> Closes: https://bugzilla.kernel.org/show_bug.cgi?id=221142
> Tested-by: Alex Bartz <alex@alstergee.com>
> Signed-off-by: Huang Wei <huangwei@kylinos.cn>

Reviewed-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>

> ---
> Changes in v2:
> - Add Tested-by tag from Alex Bartz, who verified the quirk on his
>   machine (out-of-tree build, runtime module swap, PD charging held
>   across the reload).
> 
> Changes in v3:
> - Use dev_err() instead of dev_info() for the skip message, per Greg
>   Kroah-Hartman and Heikki Krogerus, so the issue stays visible until
>   the firmware is fixed.
> - Scope the quirk to the BIOS versions confirmed broken (Q7CN44WW,
>   Q7CN78WW) instead of the whole platform, per Greg Kroah-Hartman, so
>   a future firmware that fixes UCSI is not wrongly skipped. New broken
>   versions can be added to the table as they are confirmed.
> ---
>  drivers/usb/typec/ucsi/ucsi_acpi.c | 35 ++++++++++++++++++++++++++++++
>  1 file changed, 35 insertions(+)
> 
> diff --git a/drivers/usb/typec/ucsi/ucsi_acpi.c b/drivers/usb/typec/ucsi/ucsi_acpi.c
> index 18286d3e9cc5..2708b5f90718 100644
> --- a/drivers/usb/typec/ucsi/ucsi_acpi.c
> +++ b/drivers/usb/typec/ucsi/ucsi_acpi.c
> @@ -173,6 +173,35 @@ static const struct dmi_system_id ucsi_acpi_quirks[] = {
>  	{ }
>  };
>  
> +/*
> + * Platforms with a completely broken UCSI implementation in firmware.
> + * Do not register UCSI at all on these machines.
> + *
> + * The Lenovo Legion Pro 7 16IAX10H (DMI product name 83F5) carries
> + * broken UCSI firmware in BIOS versions Q7CN44WW and Q7CN78WW; the
> + * reporter confirmed the issue is unaffected by the BIOS update
> + * between them. Match only the BIOS versions known to be broken so
> + * that a future firmware that fixes UCSI is not wrongly skipped; add
> + * new versions here as they are confirmed.
> + */
> +static const struct dmi_system_id ucsi_acpi_ignore[] = {
> +	{
> +		.matches = {
> +			DMI_MATCH(DMI_SYS_VENDOR, "LENOVO"),
> +			DMI_MATCH(DMI_PRODUCT_NAME, "83F5"),
> +			DMI_MATCH(DMI_BIOS_VERSION, "Q7CN44WW"),
> +		},
> +	},
> +	{
> +		.matches = {
> +			DMI_MATCH(DMI_SYS_VENDOR, "LENOVO"),
> +			DMI_MATCH(DMI_PRODUCT_NAME, "83F5"),
> +			DMI_MATCH(DMI_BIOS_VERSION, "Q7CN78WW"),
> +		},
> +	},
> +	{ }
> +};
> +
>  static void ucsi_acpi_notify(acpi_handle handle, u32 event, void *data)
>  {
>  	struct ucsi_acpi *ua = data;
> @@ -199,6 +228,12 @@ static int ucsi_acpi_probe(struct platform_device *pdev)
>  	if (adev->dep_unmet)
>  		return -EPROBE_DEFER;
>  
> +	if (dmi_check_system(ucsi_acpi_ignore)) {
> +		dev_err(&pdev->dev,
> +			"UCSI implementation in firmware is broken, ignoring\n");
> +		return -ENODEV;
> +	}
> +
>  	ua = devm_kzalloc(&pdev->dev, sizeof(*ua), GFP_KERNEL);
>  	if (!ua)
>  		return -ENOMEM;
> -- 
> 2.25.1

-- 
heikki