Re: [REGRESSION][BISECTED] HDA controller MMIO unreadable on Intel 5-series PCH (Ibex Peak) since v6.16

[Mario Limonciello <superm1@kernel.org>]

On 5/29/26 20:26, Filippe Duke wrote:
> #regzbot introduced: 4d4c10f763d7
> 
> Hi,
> 
> The HDA controller on old Intel 5-series PCH (Ibex  Peak) becomes non- 
> functional since 6.16.
> PCI configuration space is ok;
> PCI function power state appears to be D0;
> However MMIO always reads 0xFFFF, causing CORB timeouts;
> ALC269VB codec enumeration fails (no AFG/MFG node found).
> HW works on v6.15.x and earlier and works under WinPE on the same machine.
> 
> Bisected to first bad commit:
> 4d4c10f763d7 PCI: Explicitly put devices into D0 when initializing
> 
> On top of v6.16, commenting out the single call to
> |pci_pm_power_up_and_verify_state(dev)| in |pci_pm_init()| (drivers/pci/ 
> pci.c:3273)
> is sufficient to restore working audio. The one-line diff is attached.
> 
> HW:
> ---
> DMI:  Acer Aspire 4745G / JV41_CP, BIOS V1.25 03/16/2011 (Insyde)
> CPU: Intel Core i7 M 640
> PCH: Intel 5-series (Ibex Peak)
> HDA: 0000:00:1b.0 [8086:3b56] rev 05, SSID [1025:0357]
> Codec: Realtek ALC269VB
> GPU: Switchable:  Intel HD (Ironlake, 8086:0046) +
> AMD Radeon HD Cedar (1002:68e0, 01:00.0)
> 
> Symptom (dmesg):
> ---
> snd_hda_intel 0000:00:1b.0: enabling device (0000 -> 0002)
> snd_hda_intel 0000:00:1b.0: bound 0000:00:02.0 (ops 
> intel_audio_component_bind_ops [i915])
> snd_hda_intel 0000:00:1b.0: number of I/O streams is 30, forcing 
> separate stream tags
> snd_hda_intel 0000:00:1b.0: CORB reset timeout#2, CORBRP = 65535
> hdaudio hdaudioC0D0: no AFG or MFG node found
> hdaudio hdaudioC0D1: no AFG or MFG node found
> hdaudio hdaudioC0D2: no AFG or MFG node found
> hdaudio hdaudioC0D3: no AFG or MFG node found
> snd_hda_intel 0000:00:1b.0: no codecs initialized
> 
> Hence the sound device is missing from ALSA.
> HDMI audio on AMD (01:00.1) is not affected by the issue.
> Attaching you lshw, lspci, DSDT, dmesg and diff.
> 
> Best regards
> 
> -- 
> 
> Filippe Duke
> System Engineer
> 
> GREENNET
> Kyiv, 7B Lesi Ukrainky St.,
> Cell.: +380672772879
> Email: fduke@greennet.ua
> Web: https://greennet.ua
> 

There have been a few commits that have adjusted the behavior due to the 
commit you identified.  So I believe it's potentially the tip of the 
iceberg.

Can you please see if this issue still exists on 7.1-rc5?

This is the most recent commit (from 7.1-rc3) that I know of changing it.

https://git.kernel.org/torvalds/c/909f7bf9b080c

Thanks!