Re: [PATCH] PCI: Override Intel Apollo Lake audio class on Gauteng W3S laptops [8086:5a98]

From: Pierre-Louis Bossart

Date: Thu Nov 19 2026 - 05:41:35 EST


On 11/19/26 01:52, Steven 'Steve' Kendall wrote:
> Gauteng Education Dept MODEL-W3S laptops advertise PCI class 0x040300
> to use the legacy HDA driver rather than the Audio DSP on the Intel
> Apollo Lake audio controller ([8086:5a98]). However, this board uses an
> I2S ES8336 codec with no analog HDA codec, and its firmware otherwise
> sets up the hardware to enable the DSP.
>
> Add a PCI header fixup for [8086:5a98] on Gauteng W3S systems to
> override class 0x040300 to 0x040100 so the SOF driver can claim the
> device. Notably this fixes the internal speakers and headphones but
> not jack detection.
>
> This patch and its initial changelog were generated by Gemini during
> an interactive debugging session given dmesg, lspci and similar
> dumps from Gauteng W3S and MODEL-W3S laptops. The resulting quirk
> was reviewed and tested on physical MODEL-W3S hardware and on a W3S
> with slightly different hardware.
>
> Signed-off-by: Steven 'Steve' Kendall <skend@xxxxxxxxxxxx>
> ---
> drivers/pci/quirks.c | 34 ++++++++++++++++++++++++++++++++++
> 1 file changed, 34 insertions(+)
>
> diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c

Humm, this sort of quirks is usually in sound/hda/core/intel-dsp-config.c.

I don't think we've ever touched drivers/pci/quirks.c for DSP-related
changes, did we?

And last you could do this driver selection with the kernel parameter
snd_intel_dspcfg.dsp_driver=2.

> index 7f1f47ea2ee4..908d038fa29a 100644
> --- a/drivers/pci/quirks.c
> +++ b/drivers/pci/quirks.c
> @@ -754,6 +754,40 @@ DECLARE_PCI_FIXUP_CLASS_HEADER(PCI_VENDOR_ID_SYNOPSYS, PCI_ANY_ID,
> PCI_CLASS_SERIAL_USB_XHCI, 0,
> quirk_synopsys_haps);
>
> +/*
> + * Gauteng Education Dept MODEL-W3S laptops report PCI class 0x040300
> + * (PCI_CLASS_MULTIMEDIA_HD_AUDIO) on the Apollo Lake audio controller
> + * ([8086:5a98]) even though the Audio DSP (BAR 4), NHLT ACPI table, and I2S
> + * ESSX8336 codec are enabled and no analog HDA codec is present.
> + *
> + * Override the class to 0x040100 (PCI_CLASS_MULTIMEDIA_AUDIO << 8) so SOF can
> + * claim the device.
> + */
> +static const struct dmi_system_id gauteng_w3s_dmi_table[] = {
> + {
> + .ident = "Gauteng Education Dept W3S",
> + .matches = {
> + DMI_MATCH(DMI_SYS_VENDOR, "Gauteng Education Dept"),
> + DMI_MATCH(DMI_PRODUCT_NAME, "W3S"),
> + }
> + },
> + {}
> +};
> +
> +static void quirk_gauteng_w3s_audio_class(struct pci_dev *pdev)
> +{
> + u32 class = pdev->class;
> +
> + if (dmi_check_system(gauteng_w3s_dmi_table)) {
> + pdev->class = PCI_CLASS_MULTIMEDIA_AUDIO << 8;
> + pci_info(pdev, FW_BUG "PCI class overridden (%#08x -> %#08x) so SOF driver can claim this instead of snd_hda_intel\n",
> + class, pdev->class);
> + }

also is SOF really used for an ApolloLake product? I thought all
SOF-based Chromebooks started with GeminiLake?