Re: Linux 6.1 and 6.2-rc make mousewheel on Logitech G903 (046d:c091) report too many non-hires events

[Salvatore Bonaccorso <carnil@debian.org>]

On Wed, Jan 25, 2023 at 11:01:00PM +0100, Tobias Klausmann wrote:
> Hi!
> 
> As it says in the subject.
> 
> At some point between 6.0 and 6.1, the kernel (if HID_LOGITECH_HIDPP was
> m or y) started reporting a full event for every hires event on a
> Logitech G903. 
> 
> In my quest of finding out what the root cause is, ive filed these bugs:
> 
> https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1029584
> https://gitlab.freedesktop.org/libinput/libinput/-/issues/852
> 
> The libinput one has the most detail. The most relevant one is this log
> from libinput record, showing what happens if I turn the mousewheel by
> one notch:
> 
>     events:
>     # Current time is 20:37:11
>     - evdev:
>       - [  3,  63476,   2,   8,      -1] # EV_REL / REL_WHEEL                -1
>       - [  3,  63476,   2,  11,    -120] # EV_REL / REL_WHEEL_HI_RES       -120
>       - [  3,  63476,   0,   0,       0] # ------------ SYN_REPORT (0) ---------- +3063ms
>     - evdev:
>       - [  3, 139476,   2,   8,      -1] # EV_REL / REL_WHEEL                -1
>       - [  3, 139476,   2,  11,    -120] # EV_REL / REL_WHEEL_HI_RES       -120
>       - [  3, 139476,   0,   0,       0] # ------------ SYN_REPORT (0) ---------- +76ms
>     - evdev:
>       - [  3, 197482,   2,   8,      -1] # EV_REL / REL_WHEEL                -1
>       - [  3, 197482,   2,  11,    -120] # EV_REL / REL_WHEEL_HI_RES       -120
>       - [  3, 197482,   0,   0,       0] # ------------ SYN_REPORT (0) ---------- +58ms
>     - evdev:
>       - [  3, 227485,   2,   8,      -1] # EV_REL / REL_WHEEL                -1
>       - [  3, 227485,   2,  11,    -120] # EV_REL / REL_WHEEL_HI_RES       -120
>       - [  3, 227485,   0,   0,       0] # ------------ SYN_REPORT (0) ---------- +30ms
>     - evdev:
>       - [  3, 240484,   2,   8,      -1] # EV_REL / REL_WHEEL                -1
>       - [  3, 240484,   2,  11,    -120] # EV_REL / REL_WHEEL_HI_RES       -120
>       - [  3, 240484,   0,   0,       0] # ------------ SYN_REPORT (0) ---------- +13ms
>     - evdev:
>       - [  3, 250470,   2,   8,      -1] # EV_REL / REL_WHEEL                -1
>       - [  3, 250470,   2,  11,    -120] # EV_REL / REL_WHEEL_HI_RES       -120
>       - [  3, 250470,   0,   0,       0] # ------------ SYN_REPORT (0) ---------- +10ms
>     - evdev:
>       - [  3, 256486,   2,   8,      -1] # EV_REL / REL_WHEEL                -1
>       - [  3, 256486,   2,  11,    -120] # EV_REL / REL_WHEEL_HI_RES       -120
>       - [  3, 256486,   0,   0,       0] # ------------ SYN_REPORT (0) ---------- +6ms
>     - evdev:
>       - [  3, 264472,   2,   8,      -1] # EV_REL / REL_WHEEL                -1
>       - [  3, 264472,   2,  11,    -120] # EV_REL / REL_WHEEL_HI_RES       -120
>       - [  3, 264472,   0,   0,       0] # ------------ SYN_REPORT (0) ---------- +8ms
>     # Current time is 20:37:15
> 
> Note how it reports both REL_WHEEL and REL_WHEEL_HI_RES at every
> instance.
> 
> I've tried assorted kernels, both Debian as well as hand-compiled
> vanilla ones, and the problem only shows up if hid_logitech_hidpp.ko is
> loaded, so I am blacklisting it for now. Also note that even unloading
> the module didn't fix the behavior, though that may be because X11 was
> still runing and so libinput's state ight have been still mangled.
> 
> The most recent kernels I have tested with are 6.1.4 and 6.2-rc5.
> 
> Best,
> Tobias

Let's loop in as well the regresssions list.

#regzbot introduced: v6.0..v6.1
#regzbot link: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1029584
#regzbot link: https://gitlab.freedesktop.org/libinput/libinput/-/issues/852

Regards,
Salvatore