[Linux-kernel-mentees] Advice on fixing the bug of MSFT0001:00 04F3:Touchpad being handled by hid_multitouch by mistake

Benjamin Tissoires benjamin.tissoires at redhat.com
Fri Sep 4 16:47:42 UTC 2020

On Fri, Sep 4, 2020 at 4:59 PM Coiby Xu <coiby.xu at gmail.com> wrote:
> On Fri, Sep 04, 2020 at 10:16:51AM +0200, Benjamin Tissoires wrote:
> >Hi,
> >
> >On Tue, Aug 11, 2020 at 11:15 AM Coiby Xu <coiby.xu at gmail.com> wrote:
> >>
> >> Hi,
> >>
> >> I'm working on a touchpad device issue as reported on
> >> https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1887190.
> >>
> >> This touchpad device MSFT0001:00 04F3:Touchpad should be handled by
> >> hid_rmi. But currently hid-core.c chooses hid_multitouch by mistake,
> >>
> >>      1. When scanning this device's report descriptor, HID_DG_CONTACTID
> >>         usage is found. Thus group HID_GROUP_MULTITOUCH is assigned to
> >>         the device.
> >>      2. The flag HID_SCAN_FLAG_MT_WIN_8 is also found. Thus group
> >>         HID_GROUP_MULTITOUCH_WIN_8 is assigned to the device.
> >>      3. hid-multitouch.c claims handling devices with the group of
> >>
> >>          static const struct hid_device_id mt_devices[] = {
> >>                 /* Generic MT device */
> >>
> >>                 /* Generic Win 8 certified MT device */
> >>                 {  .driver_data = MT_CLS_WIN_8,
> >>                         HID_DEVICE(HID_BUS_ANY, HID_GROUP_MULTITOUCH_WIN_8,
> >>                                 HID_ANY_ID, HID_ANY_ID) },
> >>                 { }
> >>          };
> >>
> >> There are several potential solutions,
> >>      - Let the device vendor fix this problem since this device's report
> >>        descriptor shouldn't have the HID_DG_CONTACTID usage.
> >>      - Make it a special case by setting the device's group to
> >>        HID_GROUP_RMI in hid_scan_report when vendor id and product ID
> >>        are matched.
> >>      - hid-quirks.c seems to be designed to handle special cases, is it
> >>        suitable for this case?
> >
> >AFAIU, the touchpad doesn't work at all with hid-multitouch. So I
> >guess the best is to add the VID/PID to hid-quirks.c in
> >hid_have_special_driver[], and add it to the hid-rmi driver too.
> >This way, you will ensure hid-rmi will pick up the device all the time.
> >
> >Cheers,
> >Benjamin
> Thank you for the advice! I have exactly adopted this approach by looking
> at commit e9287099ba6539bccb20cd791269186f3ae28b85
> ("HID: rmi: Add support for the touchpad in the Razer Blade 14 laptop")
> as an example.
> My previous email is a bit misleading because 0x04F3 is the vendor code
> of ELAN while hid-rmi is for the Synaptics touchpad. And actually this
> laptop model of Lenovo Legion-5 15ARH05 is shipped with both kinds of
> touchpads,
> - for the Synaptics touchpad, hid-rmi could almost handle it perfectly
>    except the clicking is not sensitive enough. I need to let my finger
>    linger on the touchpad for a while. I notice when I click on the touchpad,
>    an HID report would be received by hid-recorder. But evtest couldn't receive
>    any EV_ event. If hid-multitouch is handling the device, the cursor
>    won't move but 2-4 finger touching events could still be received by
>    evtest.
> - for the ELAN touchpad, only HID reporters can be read and parsed by
>    hid-core then two input devices (mouse and touchpad) could created by
>    hid-multitouch as seen from /proc/bus/input/devices. But hid-recorder
>    could never get any HID report.

huh. So in both cases you have a buggy touchpad with hid-multitouch :(

Do both touchpads share the same VID/PID? If so, this is going to be
interesting to decide how any of those touchpad should be handled.

> AFAIU, isn't hid-multitouch supposed to be the implementation of Windows
> Precision Touchpad?

yes, it works for most of those. The only ones that are not working
are usually because OEM or device makers tend to do "fun" things.

>  And since Precision touchpad is mandatory for
> Windows 10 notebooks and this laptop model of Lenovo Legion-5 15ARH05
> seem to have been certificated by Windows 10, does it mean by theory
> hid-multiouch could handle these two touchpad devices?

Well, it should, yes, but it clearly can not. You can try to give a
try at hid-recorder from
https://gitlab.freedesktop.org/libevdev/hid-tools. This will show what
is actually exported by the hardware before any processing by the
kernel. Maybe there is a new simple thing to do in hid-multitouch to
handle those devices.

Also, last time I heard, Synaptics dropped the use of hid-rmi in favor
of hid-multitouch. Any hid-rmi touchpad should be able to use
hid-multitouch, as this is the preferred way on Windows. But sometimes
the various teams decide to change the rules.

> Anyway, it seems I need to install Windows Driver Kit to capture&analyze
> HID reports to see what's happening. Or do you have any suggestion?

Unless it changed recently (I think I have seen something like that
recently), I had to write a I2C man-in-the-middle to dump the logs
from Windows. Project is at
https://github.com/bentiss/SimplePeripheralBusProbe and requires a
little bit of manual work to be able to start capturing data :(


> --
> Best regards,
> Coiby

More information about the Linux-kernel-mentees mailing list