Re: [PATCH 2/2] platform/x86: msi-wmi: Add MSI Claw M-Center keys
From: Armin Wolf
Date: Fri Jun 12 2026 - 17:47:18 EST
Am 12.06.26 um 20:25 schrieb Derek John Clark:
On Fri, Jun 12, 2026 at 10:54 AM Armin Wolf <W_Armin@xxxxxx> wrote:
Am 12.06.26 um 00:39 schrieb Derek J. Clark:
MSI Claw devices produce WMI events through the MSI WMI hotkeys GUID for
some of their buttons. When pressed, these cause spam in the kernel. For
the majority of devices these events can be safely ignored as they are
duplicated by the AT Translated Set 2 Keyboard device exposed as an
evdev. For the MSI Claw A8 BZ2EM model's M-Center Menu button (left of
the screen) there is no associated keyboard event, so this event must be
exposed. Map this button to the same scancode produced by the AT
Keyboard device on other models. This does cause double F15 events on
the A1M, 7 AI+ A2VM, and 8 AI+ A2VM, but it appears to be harmless in my
testing.
Signed-off-by: Derek J. Clark <derekjohn.clark@xxxxxxxxx>
---
drivers/platform/x86/msi-wmi.c | 22 ++++++++++++++++++++++
1 file changed, 22 insertions(+)
diff --git a/drivers/platform/x86/msi-wmi.c b/drivers/platform/x86/msi-wmi.c
index d00ced756581..73a4f0d2dc04 100644
--- a/drivers/platform/x86/msi-wmi.c
+++ b/drivers/platform/x86/msi-wmi.c
@@ -46,6 +46,12 @@ enum msi_scancodes {
WIND_KEY_WLAN = 0x5f, /* Fn+F11 Wi-Fi toggle */
WIND_KEY_TURBO, /* Fn+F10 turbo mode toggle */
WIND_KEY_ECO = 0x69, /* Fn+F10 ECO mode toggle */
+ /* MSI Claw keys */
+ CLAW_KEY_VOLUMEDOWN = 0x21,
+ CLAW_KEY_CENTER = 0x29, /* MSI M-Center main menu */
+ CLAW_KEY_QUICK_LONG = 0x2a, /* MSI M-Center quick access long hold */
+ CLAW_KEY_VOLUMEUP = 0x32,
+ CLAW_KEY_QUICK_SHORT = 0x58, /* MSI M-Center quick access short press */
};
static struct key_entry msi_wmi_keymap[] = {
{ KE_KEY, MSI_KEY_BRIGHTNESSUP, {KEY_BRIGHTNESSUP} },
@@ -69,6 +75,15 @@ static struct key_entry msi_wmi_keymap[] = {
{ KE_KEY, WIND_KEY_TURBO, {KEY_PROG1} },
{ KE_KEY, WIND_KEY_ECO, {KEY_PROG2} },
+ /* These are MSI Claw keys, used for MSI M-Center in Windows */
+ { KE_KEY, CLAW_KEY_CENTER, {KEY_F15} },
+
+ /* These MSI Claw keys work without WMI. Ignore them to avoid double keycodes */
+ { KE_IGNORE, CLAW_KEY_QUICK_SHORT },
+ { KE_IGNORE, CLAW_KEY_QUICK_LONG },
+ { KE_IGNORE, CLAW_KEY_VOLUMEUP },
+ { KE_IGNORE, CLAW_KEY_VOLUMEDOWN },
+
{ KE_END, 0 }
};
@@ -183,6 +198,13 @@ static void msi_wmi_notify(union acpi_object *obj, void *context)
eventcode = obj->integer.value;
pr_debug("Eventcode: 0x%x\n", eventcode);
break;
+ case ACPI_TYPE_BUFFER:
+ if (!obj->buffer.length)
+ return;
+
+ eventcode = obj->buffer.pointer[0];
+ pr_debug("Eventcode: 0x%x\n", eventcode);
Can you share the BMOF associated with this event? I suspect that the buffer should
contain a u32.
The GUID's don't appear to contain any bmof data:
$ lsmod | grep bmof
wmi_bmof 12288 0
wmi 32768 4 video,wmi_bmof,msi_wmi,msi_wmi_platform
$ tree -l -L 2
.
├── 1E2A0DA0-2B9E-424F-9C87-B1DAC3F4E9DA-3 ->
../../../devices/pci0000:00/0000:00:02.3/0000:c3:00.0/PNP0C14:01/wmi_bus/wmi_bus-PNP0C14:01/1E2A0DA0-2B9E-424F-9C87-B1DAC3F4E9DA-3
│ ├── driver_override
│ ├── expensive
│ ├── guid
│ ├── instance_count
│ ├── modalias
│ ├── notify_id
│ ├── power
│ ├── subsystem -> ../../../../../../../../bus/wmi
│ └── uevent
├── 5B3CC38A-40D9-7245-8AE6-1145B751BE3F-1 ->
../../../devices/pci0000:00/0000:00:14.3/PNP0C09:00/PNP0C14:00/wmi_bus/wmi_bus-PNP0C14:00/5B3CC38A-40D9-7245-8AE6-1145B751BE3F-1
│ ├── driver_override
│ ├── expensive
│ ├── guid
│ ├── instance_count
│ ├── modalias
│ ├── notify_id
│ ├── power
│ ├── subsystem -> ../../../../../../../../bus/wmi
│ └── uevent
├── ABBC0F6E-8EA1-11D1-00A0-C90629100000-0 ->
../../../devices/pci0000:00/0000:00:14.3/PNP0C09:00/PNP0C14:00/wmi_bus/wmi_bus-PNP0C14:00/ABBC0F6E-8EA1-11D1-00A0-C90629100000-0
│ ├── driver -> ../../../../../../../../bus/wmi/drivers/msi-wmi-platform
│ ├── driver_override
│ ├── expensive
│ ├── guid
│ ├── hwmon
│ ├── instance_count
│ ├── modalias
│ ├── object_id
│ ├── platform-profile
│ ├── power
│ ├── subsystem -> ../../../../../../../../bus/wmi
│ └── uevent
└── B2526ED4-CB45-49FA-9230-8D2FE8AFB8EC-2 ->
../../../devices/pci0000:00/0000:00:02.3/0000:c3:00.0/PNP0C14:01/wmi_bus/wmi_bus-PNP0C14:01/B2526ED4-CB45-49FA-9230-8D2FE8AFB8EC-2
├── driver_override
├── expensive
├── guid
├── instance_count
├── modalias
├── object_id
├── power
├── subsystem -> ../../../../../../../../bus/wmi
└── uevent
16 directories, 28 files
I extracted the _WED event method from the DSDT. I don't have a lot of
experience with interpreting ACPI, but my understanding is that it
returns u8[2] with index 0 being the keycode and index 1 being the
state. Since we don't get events on release, that is essentially
unused. I can do a .length != 2 in the validation step before I
extract the keycode if desired.
Oh right, i forgot that the MSI software contains the necessary BMOF data xd.
Anyway, here it is for the WIND event:
[WMI, Dynamic, Provider("WmiProv"), Local("MS\\0x409"), Description("Event defined by MSI"), guid("{5B3CC38A-40D9-7245-8AE6-1145B751BE3F}")]
class MSI_Event : WMIEvent {
[key, read] string InstanceName;
[read] boolean Active;
[WmiDataId(1), read, write, Description("Event defined by MSI")] uint32 MSIEvt;
};
So the event payload is indeed a plain u32. For some reason the Windows WMI stack accepts 2 byte buffers even when at least 4 are necessary. I have no idea why Windows does accept undersized data in this case, but
i suggest you do the same.
So a length < 2 check (not !=, we need to still accept oversized buffers) like you described earlier should be enough. Just drop a comment somewhere that explains the meaning of index 1.
Thanks,
Armin Wolf
Method (_WED, 1, NotSerialized) // _Wxx: Wake Event, xx=0x00-0xFF
{
If ((Arg0 == 0xC0))
{
Name (EVRT, Buffer (0x02)
{
0x00, 0x00 // ..
})
Name (TMPB, Zero)
TMPB = WMRD (0xD8)
EVRT [Zero] = TMPB /* \_SB_.PCI0.SBRG.EC__.SCM0._WED.TMPB */
If ((TMPB == 0x08))
{
EVRT [One] = ((WMRD (0xE4) >> 0x03) & 0x1F
)
}
If ((TMPB == 0x56))
{
EVRT [One] = WMRD (0x2E)
}
If ((TMPB == 0x57))
{
EVRT [One] = WMRD (0x2E)
}
If ((TMPB == 0x5F))
{
EVRT [One] = WMRD (0x2E)
}
If ((TMPB == 0x60))
{
EVRT [One] = (WMRD (0xE4) & 0x07)
}
If ((TMPB == 0x62))
{
EVRT [One] = WMRD (0xD1)
}
If ((TMPB == 0x63))
{
EVRT [One] = WMRD (0xD1)
}
If ((TMPB == 0x73))
{
EVRT [One] = ((WMRD (0xE4) >> 0x03) & 0x1F
)
}
If ((TMPB == 0x79))
{
EVRT [One] = ((WMRD (0xE4) >> 0x03) & 0x1F
)
}
If ((TMPB == 0x7B))
{
EVRT [One] = WMRD (0x2E)
}
If ((TMPB == 0x84))
{
EVRT [One] = WMRD (0x2E)
}
If ((TMPB == 0x02))
{
EVRT [One] = ((WMRD (0xE4) >> 0x03) & 0x1F
)
}
WMWT (0xD8, Zero)
Return (EVRT) /* \_SB_.PCI0.SBRG.EC__.SCM0._WED.EVRT */
}
Return (Zero)
}
Also the _WDG that points the Wind GUID to arg 0xC0:
Name (_WDG, Buffer (0x64)
{
/* 0000 */ 0x6A, 0x0F, 0xBC, 0xAB, 0xA1, 0x8E, 0xD1,
0x11, // j.......
/* 0008 */ 0x00, 0xA0, 0xC9, 0x06, 0x29, 0x10, 0x00,
0x00, // ....)...
/* 0010 */ 0x41, 0x4B, 0x01, 0x01, 0x6B, 0x0F, 0xBC,
0xAB, // AK..k...
/* 0018 */ 0xA1, 0x8E, 0xD1, 0x11, 0x00, 0xA0, 0xC9,
0x06, // ........
/* 0020 */ 0x29, 0x10, 0x00, 0x00, 0x41, 0x4C, 0x01,
0x01, // )...AL..
/* 0028 */ 0x6D, 0x0F, 0xBC, 0xAB, 0xA1, 0x8E, 0xD1,
0x11, // m.......
/* 0030 */ 0x00, 0xA0, 0xC9, 0x06, 0x29, 0x10, 0x00,
0x00, // ....)...
/* 0038 */ 0x41, 0x4A, 0x01, 0x02, 0x6E, 0x0F, 0xBC,
0xAB, // AJ..n...
/* 0040 */ 0xA1, 0x8E, 0xD1, 0x11, 0x00, 0xA0, 0xC9,
0x06, // ........
/* 0048 */ 0x29, 0x10, 0x00, 0x00, 0x41, 0x4D, 0x01,
0x02, // )...AM..
/* 0050 */ 0x8A, 0xC3, 0x3C, 0x5B, 0xD9, 0x40, 0x45,
0x72, // ..<[.@Er
/* 0058 */ 0x8A, 0xE6, 0x11, 0x45, 0xB7, 0x51, 0xBE,
0x3F, // ...E.Q.?
/* 0060 */ 0xC0, 0x00, 0x01, 0x09
// ....
})
...
Thanks,
Derek
Thanks,
Armin Wolf
+ break;
default:
pr_info("Unknown event received\n");
return;