[BUG] Input: atkbd: REDMI Book Pro 14 2026 (TM2426) needs dumbkbd workaround

[BUG] Input: atkbd: REDMI Book Pro 14 2026 (TM2426) needs dumbkbd workaround

[Jame Jao]

Hi Dmitry and linux-input,

I would like to report a built-in keyboard compatibility issue on the
Xiaomi REDMI Book Pro 14 2026 (TM2426).

Original failure, observed on Arch Linux 7.2.8-zen1-2-zen:
The internal keyboard could not be used after Linux booted; an external
USB keyboard was needed. Adding i8042.dumbkbd=1 made it usable.

Current verification (2026-10-17):
The machine still boots with i8042.dumbkbd=1. The built-in device is
"AT Translated Set 2 keyboard", isa0060/serio0/input0, with EV=100013
and no EV_LED capability. There is no input Caps Lock LED interface.
I understand the missing LED interface is an expected consequence of
dumbkbd, rather than a separate LED regression. A local userspace
workaround writes SETLEDS through /dev/port; it is not an input driver fix.

Hardware:
  sys_vendor: XIAOMI
  product_name: REDMI Book Pro 14 2026
  board_name: TM2426
  BIOS: RMAPT4B0P0909, 05/21/2026
  EC firmware: 1.9
  CPU: Intel Core Ultra 5 338H

Current kernel:
  Linux 7.2.9-zen1-1.1-zen, x86_64
  Arch package linux-zen 7.2.9.zen1-1.1

This is a locally built zen kernel with a small bitland-mifs-wmi
suspend/resume error-handling change only. No atkbd/i8042 or graphics
change was made in that local patch. Kernel taint is 576 (U and W);
unsafe xe options and an intel_bios_init warning are also present.

The current upstream atkbd.c DMI table, inspected on 2026-10-17, has
atkbd_deactivate_fixup entries for Xiaomi Book Pro 14 (TM2424) and
REDMI Book Pro 16 2026 (TM2425), but no matching product entry for this
14-inch REDMI model. This suggests the same quirk might be applicable,
but that is only a hypothesis: I have NOT tested that change on TM2426.

Source inspected:
https://kernel.googlesource.com/pub/scm/linux/kernel/git/torvalds/linux/+/refs/heads/master/drivers/input/keyboard/atkbd.c

Limitations:
I have not repeated a boot without dumbkbd during this reporting session,
and have not reproduced the original failure on current mainline or
drm-tip. The original no-workaround failure above is historical; the
current device capability and DMI information were checked again today.
Please do not treat this as a tested patch or a Tested-by statement.

Could you advise whether testing atkbd_deactivate_fixup for this DMI
product is the appropriate next step, and what diagnostic data would
be useful?

Thanks,
Eason

Re: [BUG] Input: atkbd: REDMI Book Pro 14 2026 (TM2426) needs dumbkbd workaround

[Dmitry Torokhov]

Hi Jame,

On Fri, Oct 16, 2026 at 01:41:19PM -0500, Jame Jao wrote:
> Hi Dmitry and linux-input,
> 
> I would like to report a built-in keyboard compatibility issue on the
> Xiaomi REDMI Book Pro 14 2026 (TM2426).
> 
> Original failure, observed on Arch Linux 7.2.8-zen1-2-zen:
> The internal keyboard could not be used after Linux booted; an external
> USB keyboard was needed. Adding i8042.dumbkbd=1 made it usable.
> 
> Current verification (2026-10-17):
> The machine still boots with i8042.dumbkbd=1. The built-in device is
> "AT Translated Set 2 keyboard", isa0060/serio0/input0, with EV=100013
> and no EV_LED capability. There is no input Caps Lock LED interface.
> I understand the missing LED interface is an expected consequence of
> dumbkbd, rather than a separate LED regression. A local userspace
> workaround writes SETLEDS through /dev/port; it is not an input driver fix.
> 
> Hardware:
>   sys_vendor: XIAOMI
>   product_name: REDMI Book Pro 14 2026
>   board_name: TM2426
>   BIOS: RMAPT4B0P0909, 05/21/2026
>   EC firmware: 1.9
>   CPU: Intel Core Ultra 5 338H
> 
> Current kernel:
>   Linux 7.2.9-zen1-1.1-zen, x86_64
>   Arch package linux-zen 7.2.9.zen1-1.1
> 
> This is a locally built zen kernel with a small bitland-mifs-wmi
> suspend/resume error-handling change only. No atkbd/i8042 or graphics
> change was made in that local patch. Kernel taint is 576 (U and W);
> unsafe xe options and an intel_bios_init warning are also present.
> 
> The current upstream atkbd.c DMI table, inspected on 2026-10-17, has
> atkbd_deactivate_fixup entries for Xiaomi Book Pro 14 (TM2424) and
> REDMI Book Pro 16 2026 (TM2425), but no matching product entry for this
> 14-inch REDMI model. This suggests the same quirk might be applicable,
> but that is only a hypothesis: I have NOT tested that change on TM2426.
> 
> Source inspected:
> https://kernel.googlesource.com/pub/scm/linux/kernel/git/torvalds/linux/+/refs/heads/master/drivers/input/keyboard/atkbd.c
> 
> Limitations:
> I have not repeated a boot without dumbkbd during this reporting session,
> and have not reproduced the original failure on current mainline or
> drm-tip. The original no-workaround failure above is historical; the
> current device capability and DMI information were checked again today.
> Please do not treat this as a tested patch or a Tested-by statement.
> 
> Could you advise whether testing atkbd_deactivate_fixup for this DMI
> product is the appropriate next step, and what diagnostic data would
> be useful?

Yes, can you please try adding a quirk similar to TM2425 and let us know
if this solves the issue for you? Then we can include it in the kernel.

Thanks.

-- 
Dmitry

[PATCH] Input: atkbd - skip deactivate for Xiaomi Redmi Book Pro 14 2026

[Jame Jao]

[-- Attachment #1: Type: text/plain, Size: 5749 bytes --]

Hi Dmitry,

thanks, the quirk works.  Patch below; it is also attached as
atkbd-tm2426.patch in case anything gets wrapped on the way.

Test setup: Xiaomi Redmi Book Pro 14 2026, board TM2426, BIOS RMAPT4B0P0909
(2026-05-21).  Arch linux-zen 7.2.9 build with this patch on top; the only
other local change in that build is an unrelated platform/x86
suspend/resume patch, input code is otherwise untouched.

With i8042.dumbkbd=1 removed from the kernel command line and the quirk
active:

  uname -r        7.2.9-zen1-1.2-zen

  /proc/bus/input/devices
  N: Name="AT Translated Set 2 keyboard"
  P: Phys=isa0060/serio0/input0
  H: Handlers=sysrq kbd leds event3
  B: EV=120013          (EV_LED is present; dumbkbd used to give EV=100013)

  /sys/class/leds
  input3::capslock  input3::numlock  input3::scrolllock

  writing input3::capslock/brightness 1 then 0 round-trips, and the Caps
  Lock lamp follows the key again.

The built-in keyboard produced 127 key events with 34 distinct key codes
during the test, so it really is usable without the workaround.

Minor observation: the two "Spurious ACK" lines in dmesg come from an old
local userspace helper that pokes /dev/port for the LED; that helper is
redundant now and I am removing it.

From: Eason <jaojame6@gmail.com>
Subject: [PATCH] Input: atkbd - skip deactivate for Xiaomi Redmi Book
Pro 14 2026

The built-in keyboard of the Xiaomi Redmi Book Pro 14 2026 (board TM2426)
types in GRUB but stops delivering keystrokes once the kernel has booted.
The same failure mode is already handled for the sibling machines
"Xiaomi Book Pro 14" (TM2424) and "REDMI Book Pro 16 2026" (TM2425) by
skipping ATKBD_CMD_RESET_DIS in atkbd_probe(), which the embedded
controller that emulates the PS/2 keyboard does not survive.

Add the TM2426 DMI entry with the same callback.  Testing without
i8042.dumbkbd=1 on this machine, the built-in keyboard delivers key
events again, and the driver now registers the LED class devices that
dumbkbd used to remove (input3::capslock, input3::numlock,
input3::scrolllock).  Writing input3::capslock/brightness round-trips and
the Caps Lock lamp follows the key again.

Tested-by: Eason <jaojame6@gmail.com>
Signed-off-by: Eason <jaojame6@gmail.com>
---
 drivers/input/keyboard/atkbd.c | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/drivers/input/keyboard/atkbd.c b/drivers/input/keyboard/atkbd.c
--- a/drivers/input/keyboard/atkbd.c
+++ b/drivers/input/keyboard/atkbd.c
@@ -1962,6 +1962,14 @@ static const struct dmi_system_id
atkbd_dmi_quirk_table[] __initconst = {
 		},
 		.callback = atkbd_deactivate_fixup,
 	},
+	{
+		/* Xiaomi Redmi Book Pro 14 2026 (TM2426) */
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "XIAOMI"),
+			DMI_MATCH(DMI_PRODUCT_NAME, "REDMI Book Pro 14 2026"),
+		},
+		.callback = atkbd_deactivate_fixup,
+	},
 	{ }
 };

-- 
2.51.0

Thanks,
Eason


On Fri, 16 Oct 2026 12:02:21 -0700, Dmitry Torokhov
<dmitry.torokhov@gmail.com> wrote:
> Hi Jame,
>
> On Fri, Oct 16, 2026 at 01:41:19PM -0500, Jame Jao wrote:
> > Hi Dmitry and linux-input,
> >
> > I would like to report a built-in keyboard compatibility issue on the
> > Xiaomi REDMI Book Pro 14 2026 (TM2426).
> >
> > Original failure, observed on Arch Linux 7.2.8-zen1-2-zen:
> > The internal keyboard could not be used after Linux booted; an external
> > USB keyboard was needed. Adding i8042.dumbkbd=1 made it usable.
> >
> > Current verification (2026-10-17):
> > The machine still boots with i8042.dumbkbd=1. The built-in device is
> > "AT Translated Set 2 keyboard", isa0060/serio0/input0, with EV=100013
> > and no EV_LED capability. There is no input Caps Lock LED interface.
> > I understand the missing LED interface is an expected consequence of
> > dumbkbd, rather than a separate LED regression. A local userspace
> > workaround writes SETLEDS through /dev/port; it is not an input driver fix.
> >
> > Hardware:
> >   sys_vendor: XIAOMI
> >   product_name: REDMI Book Pro 14 2026
> >   board_name: TM2426
> >   BIOS: RMAPT4B0P0909, 05/21/2026
> >   EC firmware: 1.9
> >   CPU: Intel Core Ultra 5 338H
> >
> > Current kernel:
> >   Linux 7.2.9-zen1-1.1-zen, x86_64
> >   Arch package linux-zen 7.2.9.zen1-1.1
> >
> > This is a locally built zen kernel with a small bitland-mifs-wmi
> > suspend/resume error-handling change only. No atkbd/i8042 or graphics
> > change was made in that local patch. Kernel taint is 576 (U and W);
> > unsafe xe options and an intel_bios_init warning are also present.
> >
> > The current upstream atkbd.c DMI table, inspected on 2026-10-17, has
> > atkbd_deactivate_fixup entries for Xiaomi Book Pro 14 (TM2424) and
> > REDMI Book Pro 16 2026 (TM2425), but no matching product entry for this
> > 14-inch REDMI model. This suggests the same quirk might be applicable,
> > but that is only a hypothesis: I have NOT tested that change on TM2426.
> >
> > Source inspected:
> > https://kernel.googlesource.com/pub/scm/linux/kernel/git/torvalds/linux/+/refs/heads/master/drivers/input/keyboard/atkbd.c
> >
> > Limitations:
> > I have not repeated a boot without dumbkbd during this reporting session,
> > and have not reproduced the original failure on current mainline or
> > drm-tip. The original no-workaround failure above is historical; the
> > current device capability and DMI information were checked again today.
> > Please do not treat this as a tested patch or a Tested-by statement.
> >
> > Could you advise whether testing atkbd_deactivate_fixup for this DMI
> > product is the appropriate next step, and what diagnostic data would
> > be useful?
>
> Yes, can you please try adding a quirk similar to TM2425 and let us know
> if this solves the issue for you? Then we can include it in the kernel.
>
> Thanks.
>
> --
> Dmitry

[-- Attachment #2: atkbd-tm2426.patch --]
[-- Type: text/plain, Size: 1611 bytes --]

From: Eason <jaojame6@gmail.com>
Subject: [PATCH] Input: atkbd - skip deactivate for Xiaomi Redmi Book Pro 14 2026

The built-in keyboard of the Xiaomi Redmi Book Pro 14 2026 (board TM2426)
types in GRUB but stops delivering keystrokes once the kernel has booted.
The same failure mode is already handled for the sibling machines
"Xiaomi Book Pro 14" (TM2424) and "REDMI Book Pro 16 2026" (TM2425) by
skipping ATKBD_CMD_RESET_DIS in atkbd_probe(), which the embedded
controller that emulates the PS/2 keyboard does not survive.

Add the TM2426 DMI entry with the same callback.  Testing without
i8042.dumbkbd=1 on this machine, the built-in keyboard delivers key
events again, and the driver now registers the LED class devices that
dumbkbd used to remove (input3::capslock, input3::numlock,
input3::scrolllock).  Writing input3::capslock/brightness round-trips and
the Caps Lock lamp follows the key again.

Tested-by: Eason <jaojame6@gmail.com>
Signed-off-by: Eason <jaojame6@gmail.com>
---
 drivers/input/keyboard/atkbd.c | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/drivers/input/keyboard/atkbd.c b/drivers/input/keyboard/atkbd.c
--- a/drivers/input/keyboard/atkbd.c
+++ b/drivers/input/keyboard/atkbd.c
@@ -1962,6 +1962,14 @@ static const struct dmi_system_id atkbd_dmi_quirk_table[] __initconst = {
 		},
 		.callback = atkbd_deactivate_fixup,
 	},
+	{
+		/* Xiaomi Redmi Book Pro 14 2026 (TM2426) */
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "XIAOMI"),
+			DMI_MATCH(DMI_PRODUCT_NAME, "REDMI Book Pro 14 2026"),
+		},
+		.callback = atkbd_deactivate_fixup,
+	},
 	{ }
 };
 
-- 
2.51.0

Re: [PATCH] Input: atkbd - skip deactivate for Xiaomi Redmi Book Pro 14 2026

[Dmitry Torokhov]

Hi Jame,

On Sat, Oct 17, 2026 at 01:03:47AM -0400, Jame Jao wrote:
> Hi Dmitry,
> 
> thanks, the quirk works.  Patch below; it is also attached as
> atkbd-tm2426.patch in case anything gets wrapped on the way.
> 
> Test setup: Xiaomi Redmi Book Pro 14 2026, board TM2426, BIOS RMAPT4B0P0909
> (2026-05-21).  Arch linux-zen 7.2.9 build with this patch on top; the only
> other local change in that build is an unrelated platform/x86
> suspend/resume patch, input code is otherwise untouched.
> 
> With i8042.dumbkbd=1 removed from the kernel command line and the quirk
> active:
> 
>   uname -r        7.2.9-zen1-1.2-zen
> 
>   /proc/bus/input/devices
>   N: Name="AT Translated Set 2 keyboard"
>   P: Phys=isa0060/serio0/input0
>   H: Handlers=sysrq kbd leds event3
>   B: EV=120013          (EV_LED is present; dumbkbd used to give EV=100013)
> 
>   /sys/class/leds
>   input3::capslock  input3::numlock  input3::scrolllock
> 
>   writing input3::capslock/brightness 1 then 0 round-trips, and the Caps
>   Lock lamp follows the key again.
> 
> The built-in keyboard produced 127 key events with 34 distinct key codes
> during the test, so it really is usable without the workaround.
> 
> Minor observation: the two "Spurious ACK" lines in dmesg come from an old
> local userspace helper that pokes /dev/port for the LED; that helper is
> redundant now and I am removing it.
> 
> From: Eason <jaojame6@gmail.com>
> Subject: [PATCH] Input: atkbd - skip deactivate for Xiaomi Redmi Book
> Pro 14 2026
> 
> The built-in keyboard of the Xiaomi Redmi Book Pro 14 2026 (board TM2426)
> types in GRUB but stops delivering keystrokes once the kernel has booted.
> The same failure mode is already handled for the sibling machines
> "Xiaomi Book Pro 14" (TM2424) and "REDMI Book Pro 16 2026" (TM2425) by
> skipping ATKBD_CMD_RESET_DIS in atkbd_probe(), which the embedded
> controller that emulates the PS/2 keyboard does not survive.
> 
> Add the TM2426 DMI entry with the same callback.  Testing without
> i8042.dumbkbd=1 on this machine, the built-in keyboard delivers key
> events again, and the driver now registers the LED class devices that
> dumbkbd used to remove (input3::capslock, input3::numlock,
> input3::scrolllock).  Writing input3::capslock/brightness round-trips and
> the Caps Lock lamp follows the key again.
> 
> Tested-by: Eason <jaojame6@gmail.com>
> Signed-off-by: Eason <jaojame6@gmail.com>

Thank you for the patch but why the From and SOB does not match the
sender?

Thanks.

-- 
Dmitry