* ASoC: rt5677: Do not apply Yoga Book ACPI GPIO mappings to Samus
@ 2026-11-16 21:24 Ted Vera
2026-11-17 8:11 ` Pierre-Louis Bossart
2026-11-17 10:54 ` johannes.goede
0 siblings, 2 replies; 6+ messages in thread
From: Ted Vera @ 2026-11-16 21:24 UTC (permalink / raw)
To: linux-sound; +Cc: regressions, oder_chiou, lgirdwood, broonie
Summary:
On a Google Pixel 2 LS Chromebook (Samus, EVT2) running Fedora 45
kernel 7.3.3-300.fc45.x86_64, the RT5677 codec repeatedly fails probe
with Failed to request POW_LDO2: -517 (-EPROBE_DEFER). The codec
appears in devices_deferred, no RT5677 ALSA card is registered, and
only HDMI audio is available.
The driver’s ACPI GPIO mapping added by commit 0cc00de3dd5b maps GPIO
resource indexes 0 and 1 to reset and pow-ldo2 for every ACPI RT5677
device. Samus’s DSDT uses those resources for codec GPIOs (jack detect
and mic-present), not Yoga Book-style external reset and power GPIOs.
The driver requests the mapped GPIOs before registering the codec’s
GPIO chip, which appears to cause the deferred-probe loop.
Test and result:
I built a temporary module against the matching Fedora kernel source
that skips devm_acpi_dev_add_driver_gpios() on DMI board Samus. With
that module, the RT5677 device bound successfully, the bdw-rt5677 ALSA
card appeared, and laptop-speaker audio worked. Secure Boot remained
enabled. This strongly supports the ACPI GPIO mapping as the cause.
Suggested fix: Restrict the mapping to devices whose ACPI GPIO
resource layout matches the Yoga Book, rather than applying it to
every RT5677 ACPI device.
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: ASoC: rt5677: Do not apply Yoga Book ACPI GPIO mappings to Samus
2026-11-16 21:24 ASoC: rt5677: Do not apply Yoga Book ACPI GPIO mappings to Samus Ted Vera
@ 2026-11-17 8:11 ` Pierre-Louis Bossart
2026-11-17 14:48 ` Ted Vera
2026-11-17 10:54 ` johannes.goede
1 sibling, 1 reply; 6+ messages in thread
From: Pierre-Louis Bossart @ 2026-11-17 8:11 UTC (permalink / raw)
To: Ted Vera, linux-sound
Cc: regressions, oder_chiou, lgirdwood, broonie, Andy Shevchenko
On 11/16/26 22:24, Ted Vera wrote:
> Summary:
> On a Google Pixel 2 LS Chromebook (Samus, EVT2) running Fedora 45
> kernel 7.3.3-300.fc45.x86_64, the RT5677 codec repeatedly fails probe
> with Failed to request POW_LDO2: -517 (-EPROBE_DEFER). The codec
> appears in devices_deferred, no RT5677 ALSA card is registered, and
> only HDMI audio is available.
>
> The driver’s ACPI GPIO mapping added by commit 0cc00de3dd5b maps GPIO
> resource indexes 0 and 1 to reset and pow-ldo2 for every ACPI RT5677
> device. Samus’s DSDT uses those resources for codec GPIOs (jack detect
> and mic-present), not Yoga Book-style external reset and power GPIOs.
> The driver requests the mapped GPIOs before registering the codec’s
> GPIO chip, which appears to cause the deferred-probe loop.
>
> Test and result:
> I built a temporary module against the matching Fedora kernel source
> that skips devm_acpi_dev_add_driver_gpios() on DMI board Samus. With
> that module, the RT5677 device bound successfully, the bdw-rt5677 ALSA
> card appeared, and laptop-speaker audio worked. Secure Boot remained
> enabled. This strongly supports the ACPI GPIO mapping as the cause.
>
> Suggested fix: Restrict the mapping to devices whose ACPI GPIO
> resource layout matches the Yoga Book, rather than applying it to
> every RT5677 ACPI device.
Humm, it's a 2015 device that I used in the past without issues. It was
one of the early SOF devices... See commit bcb43fd.
Is this a commercial version or an early prototype? I don't recall what
EVT2 stands for in Chromebooks.
Also adding Andy in Cc:, he was the one who suggested the GPIO fixes.
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: ASoC: rt5677: Do not apply Yoga Book ACPI GPIO mappings to Samus
2026-11-16 21:24 ASoC: rt5677: Do not apply Yoga Book ACPI GPIO mappings to Samus Ted Vera
2026-11-17 8:11 ` Pierre-Louis Bossart
@ 2026-11-17 10:54 ` johannes.goede
2026-11-17 17:34 ` Pierre-Louis Bossart
1 sibling, 1 reply; 6+ messages in thread
From: johannes.goede @ 2026-11-17 10:54 UTC (permalink / raw)
To: Ted Vera, linux-sound, Yauhen Kharuzhy
Cc: regressions, oder_chiou, lgirdwood, broonie
Hi,
Note adding Yauhen the author of the troublesome commit to the To: list
(why was this not done in the first place ?).
On 16-Nov-26 22:24, Ted Vera wrote:
> Summary:
> On a Google Pixel 2 LS Chromebook (Samus, EVT2) running Fedora 45
> kernel 7.3.3-300.fc45.x86_64, the RT5677 codec repeatedly fails probe
> with Failed to request POW_LDO2: -517 (-EPROBE_DEFER). The codec
> appears in devices_deferred, no RT5677 ALSA card is registered, and
> only HDMI audio is available.
>
> The driver’s ACPI GPIO mapping added by commit 0cc00de3dd5b maps GPIO
> resource indexes 0 and 1 to reset and pow-ldo2 for every ACPI RT5677
> device. Samus’s DSDT uses those resources for codec GPIOs (jack detect
> and mic-present), not Yoga Book-style external reset and power GPIOs.
> The driver requests the mapped GPIOs before registering the codec’s
> GPIO chip, which appears to cause the deferred-probe loop.
>
> Test and result:
> I built a temporary module against the matching Fedora kernel source
> that skips devm_acpi_dev_add_driver_gpios() on DMI board Samus. With
> that module, the RT5677 device bound successfully, the bdw-rt5677 ALSA
> card appeared, and laptop-speaker audio worked. Secure Boot remained
> enabled. This strongly supports the ACPI GPIO mapping as the cause.
>
> Suggested fix: Restrict the mapping to devices whose ACPI GPIO
> resource layout matches the Yoga Book, rather than applying it to
> every RT5677 ACPI device.
Only adding the GPIO mappings (devm_acpi_dev_add_driver_gpios()) on
a DMI match of the yogabook sounds good to me.
DMI strings can be taken from:
drivers/platform/x86/x86-android-tablets/dmi.c
Note there are both a X90F/L models which ship with Android and
use a different BIOS, as well as X91F/L models which ship with
Windows (same hw, different BIOS).
Yauhen, I guess the ACPI mappings where for the Windows X91 model,
since on Android I expect the GPIO's to be hardcoded in the driver?
So the devm_acpi_dev_add_driver_gpios() should be only done
if there is a DMI match for the X91F/L models, re-using
the DMI match strings from:
drivers/platform/x86/x86-android-tablets/dmi.c
Regards,
Hans
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: ASoC: rt5677: Do not apply Yoga Book ACPI GPIO mappings to Samus
2026-11-17 8:11 ` Pierre-Louis Bossart
@ 2026-11-17 14:48 ` Ted Vera
0 siblings, 0 replies; 6+ messages in thread
From: Ted Vera @ 2026-11-17 14:48 UTC (permalink / raw)
To: Pierre-Louis Bossart
Cc: linux-sound, regressions, oder_chiou, lgirdwood, broonie,
Andy Shevchenko
Good morning, Pierre-Loiuis,
Not a prototype, it is the commercial release.
Thanks,
Ted
On Tue, Nov 17, 2026 at 1:12 AM Pierre-Louis Bossart
<pierre-louis.bossart@linux.dev> wrote:
>
> On 11/16/26 22:24, Ted Vera wrote:
> > Summary:
> > On a Google Pixel 2 LS Chromebook (Samus, EVT2) running Fedora 45
> > kernel 7.3.3-300.fc45.x86_64, the RT5677 codec repeatedly fails probe
> > with Failed to request POW_LDO2: -517 (-EPROBE_DEFER). The codec
> > appears in devices_deferred, no RT5677 ALSA card is registered, and
> > only HDMI audio is available.
> >
> > The driver’s ACPI GPIO mapping added by commit 0cc00de3dd5b maps GPIO
> > resource indexes 0 and 1 to reset and pow-ldo2 for every ACPI RT5677
> > device. Samus’s DSDT uses those resources for codec GPIOs (jack detect
> > and mic-present), not Yoga Book-style external reset and power GPIOs.
> > The driver requests the mapped GPIOs before registering the codec’s
> > GPIO chip, which appears to cause the deferred-probe loop.
> >
> > Test and result:
> > I built a temporary module against the matching Fedora kernel source
> > that skips devm_acpi_dev_add_driver_gpios() on DMI board Samus. With
> > that module, the RT5677 device bound successfully, the bdw-rt5677 ALSA
> > card appeared, and laptop-speaker audio worked. Secure Boot remained
> > enabled. This strongly supports the ACPI GPIO mapping as the cause.
> >
> > Suggested fix: Restrict the mapping to devices whose ACPI GPIO
> > resource layout matches the Yoga Book, rather than applying it to
> > every RT5677 ACPI device.
>
> Humm, it's a 2015 device that I used in the past without issues. It was
> one of the early SOF devices... See commit bcb43fd.
> Is this a commercial version or an early prototype? I don't recall what
> EVT2 stands for in Chromebooks.
> Also adding Andy in Cc:, he was the one who suggested the GPIO fixes.
>
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: ASoC: rt5677: Do not apply Yoga Book ACPI GPIO mappings to Samus
2026-11-17 10:54 ` johannes.goede
@ 2026-11-17 17:34 ` Pierre-Louis Bossart
2026-11-17 20:08 ` Yauhen Kharuzhy
0 siblings, 1 reply; 6+ messages in thread
From: Pierre-Louis Bossart @ 2026-11-17 17:34 UTC (permalink / raw)
To: johannes.goede, Ted Vera, linux-sound, Yauhen Kharuzhy
Cc: regressions, oder_chiou, lgirdwood, broonie, Andy Shevchenko
On 11/17/26 11:54, johannes.goede@oss.qualcomm.com wrote:
> Hi,
>
> Note adding Yauhen the author of the troublesome commit to the To: list
> (why was this not done in the first place ?).
>
>
> On 16-Nov-26 22:24, Ted Vera wrote:
>> Summary:
>> On a Google Pixel 2 LS Chromebook (Samus, EVT2) running Fedora 45
>> kernel 7.3.3-300.fc45.x86_64, the RT5677 codec repeatedly fails probe
>> with Failed to request POW_LDO2: -517 (-EPROBE_DEFER). The codec
>> appears in devices_deferred, no RT5677 ALSA card is registered, and
>> only HDMI audio is available.
>>
>> The driver’s ACPI GPIO mapping added by commit 0cc00de3dd5b maps GPIO
>> resource indexes 0 and 1 to reset and pow-ldo2 for every ACPI RT5677
>> device. Samus’s DSDT uses those resources for codec GPIOs (jack detect
>> and mic-present), not Yoga Book-style external reset and power GPIOs.
>> The driver requests the mapped GPIOs before registering the codec’s
>> GPIO chip, which appears to cause the deferred-probe loop.
>>
>> Test and result:
>> I built a temporary module against the matching Fedora kernel source
>> that skips devm_acpi_dev_add_driver_gpios() on DMI board Samus. With
>> that module, the RT5677 device bound successfully, the bdw-rt5677 ALSA
>> card appeared, and laptop-speaker audio worked. Secure Boot remained
>> enabled. This strongly supports the ACPI GPIO mapping as the cause.
>>
>> Suggested fix: Restrict the mapping to devices whose ACPI GPIO
>> resource layout matches the Yoga Book, rather than applying it to
>> every RT5677 ACPI device.
>
> Only adding the GPIO mappings (devm_acpi_dev_add_driver_gpios()) on
> a DMI match of the yogabook sounds good to me.
>
> DMI strings can be taken from:
>
> drivers/platform/x86/x86-android-tablets/dmi.c
>
> Note there are both a X90F/L models which ship with Android and
> use a different BIOS, as well as X91F/L models which ship with
> Windows (same hw, different BIOS).
>
> Yauhen, I guess the ACPI mappings where for the Windows X91 model,
> since on Android I expect the GPIO's to be hardcoded in the driver?
>
> So the devm_acpi_dev_add_driver_gpios() should be only done
> if there is a DMI match for the X91F/L models, re-using
> the DMI match strings from:
>
> drivers/platform/x86/x86-android-tablets/dmi.c
But the open is whether those GPIO mappings belong in the codec driver?
Back in 2017 (!) Andy moved the Samus mappings to the machine driver
bdw-rt5677, why not add those YogaBook mappings to the cht_rt5677
machine driver?
Looks like the addition of those mappings for the YogaBook didn't follow
the Samus precedent, now we have two different ways of doing the same
thing even though the codec is the same...
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: ASoC: rt5677: Do not apply Yoga Book ACPI GPIO mappings to Samus
2026-11-17 17:34 ` Pierre-Louis Bossart
@ 2026-11-17 20:08 ` Yauhen Kharuzhy
0 siblings, 0 replies; 6+ messages in thread
From: Yauhen Kharuzhy @ 2026-11-17 20:08 UTC (permalink / raw)
To: Pierre-Louis Bossart
Cc: johannes.goede, Ted Vera, linux-sound, regressions, oder_chiou,
lgirdwood, broonie, Andy Shevchenko, Maurizio Casciano
Hi,
Cc: Maurizio who continues my YogaBook porting work.
вт, 17 нояб. 2026 г. в 19:34, Pierre-Louis Bossart
<pierre-louis.bossart@linux.dev>:
>
> On 11/17/26 11:54, johannes.goede@oss.qualcomm.com wrote:
> > Hi,
> >
> > Note adding Yauhen the author of the troublesome commit to the To: list
> > (why was this not done in the first place ?).
> >
> >
> > On 16-Nov-26 22:24, Ted Vera wrote:
> >> Summary:
> >> On a Google Pixel 2 LS Chromebook (Samus, EVT2) running Fedora 45
> >> kernel 7.3.3-300.fc45.x86_64, the RT5677 codec repeatedly fails probe
> >> with Failed to request POW_LDO2: -517 (-EPROBE_DEFER). The codec
> >> appears in devices_deferred, no RT5677 ALSA card is registered, and
> >> only HDMI audio is available.
> >>
> >> The driver’s ACPI GPIO mapping added by commit 0cc00de3dd5b maps GPIO
> >> resource indexes 0 and 1 to reset and pow-ldo2 for every ACPI RT5677
> >> device. Samus’s DSDT uses those resources for codec GPIOs (jack detect
> >> and mic-present), not Yoga Book-style external reset and power GPIOs.
> >> The driver requests the mapped GPIOs before registering the codec’s
> >> GPIO chip, which appears to cause the deferred-probe loop.
> >>
> >> Test and result:
> >> I built a temporary module against the matching Fedora kernel source
> >> that skips devm_acpi_dev_add_driver_gpios() on DMI board Samus. With
> >> that module, the RT5677 device bound successfully, the bdw-rt5677 ALSA
> >> card appeared, and laptop-speaker audio worked. Secure Boot remained
> >> enabled. This strongly supports the ACPI GPIO mapping as the cause.
> >>
> >> Suggested fix: Restrict the mapping to devices whose ACPI GPIO
> >> resource layout matches the Yoga Book, rather than applying it to
> >> every RT5677 ACPI device.
> >
> > Only adding the GPIO mappings (devm_acpi_dev_add_driver_gpios()) on
> > a DMI match of the yogabook sounds good to me.
> >
> > DMI strings can be taken from:
> >
> > drivers/platform/x86/x86-android-tablets/dmi.c
> >
> > Note there are both a X90F/L models which ship with Android and
> > use a different BIOS, as well as X91F/L models which ship with
> > Windows (same hw, different BIOS).
> >
> > Yauhen, I guess the ACPI mappings where for the Windows X91 model,
> > since on Android I expect the GPIO's to be hardcoded in the driver?
For Android tablet (X90) I used x86-android-tablets module to create a
ACPI-independent mapping via GPIO properties, see
https://github.com/jekhor/yogabook-linux-kernel/commit/4e3c2f5f7c74305d4825bd1c031fdb3bd16edefa
,
was submitted as
https://lore.kernel.org/platform-driver-x86/20260611-x86-tablets-yogabook-v3-0-de407672e4df@gmail.com/
If GPIO mapping in the ACPI entry for RT5677 is not generic actually, then we
can use the similar approach for Yoga Book X91 also and don't use ACPI
mapping et al.
> But the open is whether those GPIO mappings belong in the codec driver?
> Back in 2017 (!) Andy moved the Samus mappings to the machine driver
> bdw-rt5677, why not add those YogaBook mappings to the cht_rt5677
> machine driver?
I actually cannot remember why. Intel's SoC subsystem has so many similar things
made in different ways... AFAIR, I didn't find any mention about other real
devices with ACPI and RT5677 so tried to apply these mappings as generic
and belonging to the rt5677 driver instead of tricky management in the
machine driver.
>
> Looks like the addition of those mappings for the YogaBook didn't follow
> the Samus precedent, now we have two different ways of doing the same
> thing even though the codec is the same...
--
Yauhen Kharuzhy
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2026-11-17 20:08 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-11-16 21:24 ASoC: rt5677: Do not apply Yoga Book ACPI GPIO mappings to Samus Ted Vera
2026-11-17 8:11 ` Pierre-Louis Bossart
2026-11-17 14:48 ` Ted Vera
2026-11-17 10:54 ` johannes.goede
2026-11-17 17:34 ` Pierre-Louis Bossart
2026-11-17 20:08 ` Yauhen Kharuzhy
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox