[PATCH] ALSA: hda/realtek: Fix speaker silence after S3 resume on Xiaomi Mi Laptop Pro 15

[PATCH] ALSA: hda/realtek: Fix speaker silence after S3 resume on Xiaomi Mi Laptop Pro 15

[Yuriy Padlyak]

The Xiaomi Mi Laptop Pro 15 (TM1905, subsystem 1d72:1905) ships with the
Realtek ALC256 codec on Intel Comet Lake PCH-LP. After S3 resume the
codec sets coefficient register 0x10 to 0x0220 instead of 0x0020 — bit 9
is erroneously set, which silences the internal speaker. Bluetooth and
HDMI audio are unaffected because they use different paths.

This is the same mechanism fixed for Clevo NJ51CU by commit edca7cc4b0ac
("ALSA: hda/realtek: Fix quirk for Clevo NJ51CU"), but the existing
ALC256_FIXUP_MIC_NO_PRESENCE_AND_RESUME also reconfigures pin 0x19 as a
front mic, which is wrong for this Xiaomi where pin 0x19 default is
0x411111f0 (disabled). Add a minimal fixup that only clears the stuck
coef bit, and add the Xiaomi SSID to the quirk table.

Verified by reading coef 0x10 with hda-verb after resume (returns
0x0220), writing 0x0020, and confirming the internal speaker resumes
output. With this fixup applied the bit is cleared on every codec init,
including post-resume.

Signed-off-by: Yuriy Padlyak <yuriypadlyak@gmail.com>
Cc: <stable@vger.kernel.org>
Tested-by: Yuriy Padlyak <yuriypadlyak@gmail.com>
---
 sound/hda/codecs/realtek/alc269.c | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)

diff --git a/sound/hda/codecs/realtek/alc269.c b/sound/hda/codecs/realtek/alc269.c
index d720565db..2ec0b9158 100644
--- a/sound/hda/codecs/realtek/alc269.c
+++ b/sound/hda/codecs/realtek/alc269.c
@@ -3390,6 +3390,19 @@ static void alc256_fixup_mic_no_presence_and_resume(struct hda_codec *codec,
 	}
 }
 
+static void alc256_fixup_xiaomi_pro15_resume(struct hda_codec *codec,
+					     const struct hda_fixup *fix,
+					     int action)
+{
+	/*
+	 * On the Xiaomi Mi Laptop Pro 15 (TM1905, SSID 1d72:1905) the ALC256
+	 * codec sets coefficient 0x10 bit 9 to 1 after S3 resume, silencing
+	 * the internal speaker. Bluetooth and HDMI audio are unaffected.
+	 * Clear the bit so the speaker keeps working across suspend cycles.
+	 */
+	alc_update_coef_idx(codec, 0x10, 1<<9, 0);
+}
+
 static void alc256_decrease_headphone_amp_val(struct hda_codec *codec,
 					      const struct hda_fixup *fix, int action)
 {
@@ -4041,6 +4054,7 @@ enum {
 	ALC256_FIXUP_SYSTEM76_MIC_NO_PRESENCE,
 	ALC233_FIXUP_NO_AUDIO_JACK,
 	ALC256_FIXUP_MIC_NO_PRESENCE_AND_RESUME,
+	ALC256_FIXUP_XIAOMI_PRO15_RESUME,
 	ALC285_FIXUP_LEGION_Y9000X_SPEAKERS,
 	ALC285_FIXUP_LEGION_Y9000X_AUTOMUTE,
 	ALC287_FIXUP_LEGION_16ACHG6,
@@ -6229,6 +6243,10 @@ static const struct hda_fixup alc269_fixups[] = {
 		.chained = true,
 		.chain_id = ALC269_FIXUP_HEADSET_MODE_NO_HP_MIC
 	},
+	[ALC256_FIXUP_XIAOMI_PRO15_RESUME] = {
+		.type = HDA_FIXUP_FUNC,
+		.v.func = alc256_fixup_xiaomi_pro15_resume,
+	},
 	[ALC287_FIXUP_LEGION_16ACHG6] = {
 		.type = HDA_FIXUP_FUNC,
 		.v.func = alc287_fixup_legion_16achg6_speakers,
@@ -7762,6 +7780,7 @@ static const struct hda_quirk alc269_fixup_tbl[] = {
 	SND_PCI_QUIRK(0x1d72, 0x1602, "RedmiBook", ALC255_FIXUP_XIAOMI_HEADSET_MIC),
 	SND_PCI_QUIRK(0x1d72, 0x1701, "XiaomiNotebook Pro", ALC298_FIXUP_DELL1_MIC_NO_PRESENCE),
 	SND_PCI_QUIRK(0x1d72, 0x1901, "RedmiBook 14", ALC256_FIXUP_ASUS_HEADSET_MIC),
+	SND_PCI_QUIRK(0x1d72, 0x1905, "Xiaomi Mi Laptop Pro 15", ALC256_FIXUP_XIAOMI_PRO15_RESUME),
 	SND_PCI_QUIRK(0x1d72, 0x1945, "Redmi G", ALC256_FIXUP_ASUS_HEADSET_MIC),
 	SND_PCI_QUIRK(0x1d72, 0x1947, "RedmiBook Air", ALC255_FIXUP_XIAOMI_HEADSET_MIC),
 	SND_PCI_QUIRK(0x1e39, 0xca14, "MEDION NM14LNL", ALC233_FIXUP_MEDION_MTL_SPK),
-- 
2.53.0

Re: [PATCH] ALSA: hda/realtek: Fix speaker silence after S3 resume on Xiaomi Mi Laptop Pro 15

[Takashi Iwai]

On Thu, 30 Apr 2026 00:09:03 +0200,
Yuriy Padlyak wrote:
> 
> The Xiaomi Mi Laptop Pro 15 (TM1905, subsystem 1d72:1905) ships with the
> Realtek ALC256 codec on Intel Comet Lake PCH-LP. After S3 resume the
> codec sets coefficient register 0x10 to 0x0220 instead of 0x0020 ― bit 9
> is erroneously set, which silences the internal speaker. Bluetooth and
> HDMI audio are unaffected because they use different paths.
> 
> This is the same mechanism fixed for Clevo NJ51CU by commit edca7cc4b0ac
> ("ALSA: hda/realtek: Fix quirk for Clevo NJ51CU"), but the existing
> ALC256_FIXUP_MIC_NO_PRESENCE_AND_RESUME also reconfigures pin 0x19 as a
> front mic, which is wrong for this Xiaomi where pin 0x19 default is
> 0x411111f0 (disabled). Add a minimal fixup that only clears the stuck
> coef bit, and add the Xiaomi SSID to the quirk table.
> 
> Verified by reading coef 0x10 with hda-verb after resume (returns
> 0x0220), writing 0x0020, and confirming the internal speaker resumes
> output. With this fixup applied the bit is cleared on every codec init,
> including post-resume.
> 
> Signed-off-by: Yuriy Padlyak <yuriypadlyak@gmail.com>
> Cc: <stable@vger.kernel.org>
> Tested-by: Yuriy Padlyak <yuriypadlyak@gmail.com>

Thanks, applied now.


Takashi