[PATCH RFC 0/2] platform/x86: Add Emdoor Intel EWMI platform driver

[Ziqiang Xia via B4 Relay <devnull+xiazq123666.gmail.com@kernel.org>]

This RFC patch series adds initial Linux support for the Intel EWMI
(\_SB.EWMI) WMI interface found on Emdoor-based Intel laptops, initially
matched strictly by DMI and BIOS version to the MECHREVO Wujie 14 2026
(WUJIE Series-Lark4-LNL, Intel Core Ultra 5 226V Lunar Lake, BIOS
EM_LNL326_V1.0.19).

Note on naming and scope:
- Community out-of-tree projects such as fewtarius/emdoor-wmi target AMD
  Emdoor platforms using different WMI GUIDs (8600ACCE-...) and AMD SMU
  ports. This Intel EWMI implementation (0B30FD67-1B62-4FCA-B10B-50DC7ADE6A39,
  0B30FD67-1B62-4FCA-B10B-50DC7ADE6DDD, C75665A3-4382-4646-ADC8-C81AE146E455)
  is unrelated in protocol and EC register layout, so the driver is named
  emdoor-intel-wmi.
- Legacy WMAA-WMCA WMI methods on this DSDT trigger AE_AML_BUFFER_LIMIT due
  to an out-of-bounds CreateField() and are deliberately unused. The driver
  only uses the clean WMCD, _WED(0xD0), and WMCC interfaces through modern
  wmi_driver / wmidev_invoke_method().

Hardware quirks handled and verified on real hardware:
1. Shared EC5D mailbox lag and hotkey collision:
   AML methods GWLS (0x03/01) and GFLS (0x03/03) time out after 20 ms on
   this BIOS and return the previous mailbox byte on the call itself. After
   invoking WMCD and waiting 100 ms, the driver reads the settled mailbox
   register directly and retries up to 3 times if concurrent _QF6/_QF9
   hotkey events overwrite EC5D in flight.
2. SPFM stale-EC5F drop race:
   AML method SPFM (0x03/0a) arms ZZTT/TIZZ before Notify(IETM, 0x88) and
   before writing EC5F, so an EC scan during the notify can consume a stale
   EC5F and return RTNC=0 without updating PPMD (or return RTNC=1). The
   driver verifies 03/09 readback and reissues the absolute mode up to 3 times.

Validation summary on Linux 7.3.4 (MECHREVO WUJIE Series-Lark4-LNL,
EM_LNL326_V1.0.19):
- 1500 readonly attribute/hwmon stress reads (battery & AC online): 0 errors
- 20 platform_profile switches + 40 idempotent writes: 0 errors
- Fn-lock, Win-lock, and auto_power_on write/readback cycles: pass
- High-frequency Fn+Esc / F8 hotkey concurrency stress & s2idle: pass

Specific RFC questions for reviewers:
1. Should fn_lock, win_lock, and auto_power_on stay as WMI device sysfs
   attributes (documented in Patch 1), or would pdx86 maintainers prefer a
   firmware-attributes class interface or moving any of them?
2. Event codes 0xA1 (F7 mic-mute hotkey, _QF8) and 0xA2 (Fn+R refresh rate,
   _QF5) are conservatively mapped to KEY_PROG2 / KEY_PROG3 for now while
   verifying whether i8042 also emits scancodes on some desktops/DEs. Is
   KEY_MICMUTE / KEY_REFRESH_RATE_TOGGLE preferred once verified, or
   separate input mapping?
3. Exact DMI_BIOS_VERSION matching is used because board EC offsets (0x5d,
   0xf5, temp_regs {0x70, 0x73, 0x72, 0x62}) are board/BIOS-specific. Is
   this strict match appropriate for initial upstreaming?

Signed-off-by: Ziqiang Xia <xiazq123666@gmail.com>
---
Ziqiang Xia (2):
      Documentation: ABI: testing: Add emdoor-intel-wmi sysfs ABI
      platform/x86: Add Emdoor Intel EWMI platform driver

 .../ABI/testing/sysfs-bus-wmi-devices-emdoor-ewmi  |  28 +
 MAINTAINERS                                        |   7 +
 drivers/platform/x86/Kconfig                       |  18 +
 drivers/platform/x86/Makefile                      |   1 +
 drivers/platform/x86/emdoor-intel-wmi.c            | 873 +++++++++++++++++++++
 5 files changed, 927 insertions(+)
---
base-commit: 5680519168f65666d2c03cc6ed8fe2b7052aee12
change-id: 20261113-emdoor-intel-wmi-rfc-2173530b3df6

Best regards,
--  
Ziqiang Xia <xiazq123666@gmail.com>