Re: Touchpad multitouch leaves ghost fingers

On Sep 24 2025, Yinon Burgansky wrote:

On Tue, Sep 23, 2025 at 7:30 PM Benjamin Tissoires [off-list ref] wrote:

quoted

Got something out with https://gitlab.freedesktop.org/libevdev/udev-hid-bpf/-/merge_requests/204
that seems to solve your case on the hid-recorder you provided.

Thank you so much!
The install command didn't work properly for me:

$ ./install.sh "*DLL0945*"
$ tree /usr/local/lib/firmware
/usr/local/lib/firmware
└── hid
    └── bpf
        └── 0011-Synaptics__DLL0945.bpf.o
$ reboot
$ sudo tree /sys/fs/bpf
/sys/fs/bpf
0 directories, 0 files

oops, yes, my bad. I forgot to put the correct group on the device
(multitouch_win_8 instead of generic). I've force pushed a new version,
so please redownload and reinstall it.

I tried to add it manually and it seems to work now :)

$ sudo udev-hid-bpf add /sys/bus/hid/devices/0018:06CB:CE26.0005
/usr/local/lib/firmware/hid/bpf/0011-Synaptics__DLL0945.bpf.o
$ sudo tree /sys/fs/bpf
/sys/fs/bpf
└── hid
    └── 0018_06CB_CE26_0005
        └── 0011-Synaptics__DLL0945_bpf
            └── synaptics_dll0945

4 directories, 1 file

But I have to do it manually every time.
Not sure what's wrong, maybe the added udev rule:

$ cat /etc/udev/rules.d/81-hid-bpf.rules
ACTION!="add|remove|bind", GOTO="hid_bpf_end"
SUBSYSTEM!="hid", GOTO="hid_bpf_end"

# We lookup the hwdb during bind to set the property, but we don't do
anything else
IMPORT{builtin}="hwdb --subsystem=hid --lookup-prefix=hid-bpf:"
#ACTION=="add", ENV{.HID_BPF}=="1",
RUN{program}+="@@BINDIR@@/udev-hid-bpf add $sys$devpath"
#MARKER
ACTION=="add", ENV{.HID_BPF}=="1",
RUN{program}+="/usr/local/bin/udev-hid-bpf add $sys$devpath"
#ACTION=="remove", ENV{.HID_BPF}=="1",
RUN{program}+="@@BINDIR@@/udev-hid-bpf remove $sys$devpath"
#MARKER
ACTION=="remove", ENV{.HID_BPF}=="1",
RUN{program}+="/usr/local/bin/udev-hid-bpf remove $sys$devpath"

LABEL="hid_bpf_end"

The udev rule is fine, the hwdb is not. If you look at
/etc/udev/hwdb.d/81-hid-bpf-testing.hwdb you'll see that it shows:

hid-bpf:hid:b0018g0001v000006CBp0000CE26
  HID_BPF_T_002=0011-Synaptics__DLL0945.bpf.o
  .HID_BPF=1

But if you run `cat /sys/bus/hid/devices/0018:06CB:CE26.*/modalias`
you'll see that you have something like:

hid:b0018g0004v000006CBp0000CE26

the 'g0004' is not correct and therefore the qutobinding doesn't work :(

this is the udevadm info (after reboot, it is sometimes hidraw3
sometimes hidraw4):

$  udevadm info --query=all --name=/dev/hidraw4
P: /devices/pci0000:00/0000:00:15.1/i2c_designware.1/i2c-1/i2c-DLL0945:00/0018:06CB:CE26.0005/hidraw/hidraw4
M: hidraw4
R: 4
J: c242:4
U: hidraw
D: c 242:4
N: hidraw4
L: 0
E: DEVPATH=/devices/pci0000:00/0000:00:15.1/i2c_designware.1/i2c-1/i2c-DLL0945:00/0018:06CB:CE26.0005/hidraw/hidraw4
E: DEVNAME=/dev/hidraw4
E: MAJOR=242
E: MINOR=4
E: SUBSYSTEM=hidraw
E: USEC_INITIALIZED=3980846
E: ID_VENDOR_FROM_DATABASE=Dell Inc
E: ID_PATH=pci-0000:00:15.1-platform-i2c_designware.1
E: ID_PATH_TAG=pci-0000_00_15_1-platform-i2c_designware_1
E: ID_FOR_SEAT=hidraw-pci-0000_00_15_1-platform-i2c_designware_1
E: TAGS=:seat:
E: CURRENT_TAGS=:seat:

Thank you for the quick fix, it seems to work great so far!

\o/.

Now I need to make the kernel patch. Ideally I'd like you to test it,
but worse case I can just replay the recording as it seems to be enough.

Cheers,
Benjamin