[RFC] bitland-mifs-wmi: optional Fullspeed and four distinct modes on RedmiBook Pro 16 2024 (TM2309)

[RFC] bitland-mifs-wmi: optional Fullspeed and four distinct modes on RedmiBook Pro 16 2024 (TM2309)

[zsanya322]

Hello,

Thank you for your work on RedmiBook and Bitland laptop support. As a
RedmiBook Pro 16 2024 user, I would like to suggest upstream support for
its four distinct firmware modes, especially an optional, explicitly
selectable Fullspeed/Geek mode. This is a proposal for consideration;
I understand that the interface and implementation need maintainer review.

My machine is XIAOMI / TM2309, BIOS RMAMT6B0P0B0B, Intel Core Ultra 5
125H with integrated Intel Arc graphics. I use EndeavourOS KDE and
Linux 7.2.8-arch1-2.

I found the ongoing per-machine ops discussion, including reports for
TM2309 and the four-mode test on TM2307:
https://lists.openwall.net/linux-kernel/2026/10/02/8
https://lists.openwall.net/linux-kernel/2026/10/02/47

With a local driver workaround scoped to my model and BIOS, WMI SET
followed by GET confirms these modes on AC power. The corresponding
Intel DTT odvp1 value also changes:

    User-facing mode    Firmware value    odvp1
    Quiet               2                 2
    Balanced            1                 0
    Turbo               3                 1
    Fullspeed / Geek    4                 4

All four controls were verified through a local GUI, including the
return from Fullspeed to Turbo. In this workaround, Turbo and Fullspeed
share the power-profiles-daemon "performance" profile, and a separate
control selects firmware mode 4. This is an experiment, not a proposed
stable userspace ABI.

Would it be possible to cover TM2309 in the ongoing series and keep
all four firmware modes independently selectable through an appropriate
kernel interface? If the four platform_profile values already discussed
are suitable, that would be preferable to maintaining a private module.
Fullspeed should be an explicit user choice, with whatever AC and
firmware capability checks are appropriate. I would also appreciate
guidance on the userspace work needed to make all four choices available
in the desktop; I understand that a kernel change alone cannot add a
fourth KDE/power-profiles-daemon button.

For context, a short AC test with eight CPU SHA256 workers for six
seconds per mode recorded MMIO PL1 limits of 45/49/75/90 W. Mean CPU
package consumption was 18.47/43.50/42.41/46.67 W respectively. These
are observations of adaptive policy, not requested fixed wattages or
a benchmark demonstrating a performance gain. The power-policy result
also depends on a separate local thermald workaround, which avoids
initializing an MSR limit from nominal PPCC on this exact machine while
retaining adaptive thermal policy. It should be reviewed separately.

No thermal protections were disabled. No new thermal-throttle events
appeared during this short test. Fan RPM is unavailable through the
tested firmware command, so I cannot claim that Fullspeed means a
verified 100% fan speed. Reboot, suspend/resume and AC/DC transitions
remain unverified, and I have not tested the exact current upstream
series; this is not a Tested-by for that series.

The collected measurements and local patches are available if helpful.
If this is already covered by pending work, please point me to the
relevant changes. Thank you for considering the suggestion and for
advising which parts belong in the driver versus userspace.

Best regards,
Aleksandr

Re: [RFC] bitland-mifs-wmi: optional Fullspeed and four distinct modes on RedmiBook Pro 16 2024 (TM2309)

[Anton Karasev]

Hi Aleksandr,

I have the same machine (TM2309, BIOS RMAMT6B0P0B0B), and your values
match mine: quiet 2, balanced 1, Turbo 3 and full speed 4, with odvp1
2, 0, 1 and 4 [1]. Milos found the same on the TM2307, which is built
on the same platform and BIOS family [2], and tested four distinct
profiles there [3]. The Turbo and full speed PL1 values you saw (75 and
90 W) are the AC maximums of the GDDV targets here as well.

As for the status: Ilpo asked that per-model entries be based on the
generalized profile handling Chris is preparing, not on the ops table
in Mingyou's v6-v9 patch 5 [4]. That series has not been posted yet. An
entry for the TM2307 and TM2309 would go on top of it, and I will Cc
you; a test on another TM2309 would help.

On full speed: the firmware accepts mode 4 on AC and on battery without
any check, but Fn+K never offers it on this model, even with the stock
140 W charger. As I wrote to Milos [5], I would suggest performance =
Turbo for the entry, since power-profiles-daemon maps its "performance"
profile to platform_profile "performance", and full speed as max-power,
which platform_profile_cycle() skips, so it is used only when chosen
explicitly -- which matches what you asked for. power-profiles-daemon
neither selects max-power nor recognises it yet (it keeps showing the
previous profile), so for now it would be chosen by writing
/sys/firmware/acpi/platform_profile directly. Whether max-power is the
right place for it is up to the maintainers.

One more thing for this model: please do not write to the kb_mode
attribute. On this firmware command 0x10 is the battery interface, and
writing "fixed" turns off the battery charge limit [1]. The patch that
hides kb_mode there is [6]; a test on your TM2309 would help. The
charge limit itself is the subject of a separate thread [7].

[1] https://lore.kernel.org/all/20260930015058.3076905-1-uselessfire@gmail.com/
[2] https://lore.kernel.org/all/CAObHBTwnv+kEaFdCp+cr31HtdSo3CaRq84T8vr-Pm75yZMTHqQ@mail.gmail.com/
[3] https://lore.kernel.org/all/CAObHBTxHQDfVH8Su1DLABtC5uUsAGBBk=KvcCOvtTs5gMGamnA@mail.gmail.com/
[4] https://lore.kernel.org/all/094577b3-7b96-72b6-4ef2-a5d1db457401@linux.intel.com/
[5] https://lore.kernel.org/all/20261008-tm2307-tm2309-entry-uselessfire@gmail.com/
[6] https://lore.kernel.org/all/20261008-bitland-kb-mode-v1-uselessfire@gmail.com/
[7] https://lore.kernel.org/all/20261008-xiaomi-charge-limit-0x10-uselessfire@gmail.com/

Thanks,
Anton Karasev