Bug 93641

Hello,

I have purchased an external numpad (A4Tech TK-5). When I use it with my desktop machine, I can reliably reproduce input issues described here. However, not all machines exhibit those issues (more on that later).

The problem is, that the numpad works (produces key events) only when I don't use it with my standard keyboard at the very same time. If I wait a short while (~1 second) when switching between these two devices (i.e. I wait 1 second between stopping typing on the standard keyboard and starting typing on the numpad, and vice versa), everything is OK. But when I start typing on both simultaneously, the numpad gets dead/unresponsive, and no longers send *any* events (no numbers, numlock switching doesn't work - the LED does not change state) until I disconnect it and reconnect it again.

I can reproduce this every single time on my desktop machine, but only when using my rear motherboard usb ports. When I use external front panel usb ports, I can't reproduce it. So it seems to be somehow related to a certain usb chipset? (If it is a race condition, maybe it has much higher probability to occur on a certain usb chipset). On the very same machine when using Windows 10, I can't reproduce the issue on any of those ports, it works flawlessly, so I support this is a software issue, not a hardware one.

I also saw similar issues with my Thinkpad X220, but I can't reproduce them consistently there or on will, it happens randomly and rarely. On ThinkPad T450s, I could not reproduce the issue at all. So this seems to be very much affected by hardware.

When the numpad gets dead, there is no message about it in the dmesg/system journal, even if I wait for a while. xev shows no events, even when I press buttons. libinput-debug-events shows no events, even when I press buttons. (Before it gets dead, both xev and libinput-debug-events of course show proper numpad key presses).

libinput-list-devices says:
Device:           KEYBOARD
Kernel:           /dev/input/event18
Group:            4
Seat:             seat0, default
Capabilities:     keyboard 
Tap-to-click:     n/a
Tap drag lock:    n/a
Left-handed:      n/a
Nat.scrolling:    n/a
Middle emulation: n/a
Calibration:      n/a
Scroll methods:   none
Click methods:    none
Disable-w-typing: n/a
Accel profiles:   n/a

Device:           KEYBOARD
Kernel:           /dev/input/event19
Group:            4
Seat:             seat0, default
Capabilities:     keyboard pointer 
Tap-to-click:     n/a
Tap drag lock:    n/a
Left-handed:      n/a
Nat.scrolling:    disabled
Middle emulation: n/a
Calibration:      n/a
Scroll methods:   none
Click methods:    none
Disable-w-typing: n/a
Accel profiles:   n/a

I'm not sure why it shows two devices (for my standard keyboard it also shows 2 devices), when debugging this I used event18 as the input device.

When using evemu-record, I noticed that sometimes events just stop coming from the numpad once it gets dead, but very often I see an endless stream of very fast scrolling events, like this:

E: 4.378433 0001 0045 0002      # EV_KEY / KEY_NUMLOCK          2
E: 4.378433 0000 0000 0001      # ------------ SYN_REPORT (1) ---------- +33ms
E: 4.411421 0001 0045 0002      # EV_KEY / KEY_NUMLOCK          2
E: 4.411421 0000 0000 0001      # ------------ SYN_REPORT (1) ---------- +33ms
E: 4.444406 0001 0045 0002      # EV_KEY / KEY_NUMLOCK          2
E: 4.444406 0000 0000 0001      # ------------ SYN_REPORT (1) ---------- +33ms
E: 4.477405 0001 0045 0002      # EV_KEY / KEY_NUMLOCK          2
E: 4.477405 0000 0000 0001      # ------------ SYN_REPORT (1) ---------- +33ms
E: 4.510417 0001 0045 0002      # EV_KEY / KEY_NUMLOCK          2
E: 4.510417 0000 0000 0001      # ------------ SYN_REPORT (1) ---------- +33ms
E: 4.543443 0001 0045 0002      # EV_KEY / KEY_NUMLOCK          2
E: 4.543443 0000 0000 0001      # ------------ SYN_REPORT (1) ---------- +33ms


or like this:

E: 3.800587 0001 004d 0002      # EV_KEY / KEY_KP6              2
E: 3.800587 0000 0000 0001      # ------------ SYN_REPORT (1) ---------- +33ms
E: 3.833587 0001 004d 0002      # EV_KEY / KEY_KP6              2
E: 3.833587 0000 0000 0001      # ------------ SYN_REPORT (1) ---------- +33ms
E: 3.866576 0001 004d 0002      # EV_KEY / KEY_KP6              2
E: 3.866576 0000 0000 0001      # ------------ SYN_REPORT (1) ---------- +33ms
E: 3.899576 0001 004d 0002      # EV_KEY / KEY_KP6              2
E: 3.899576 0000 0000 0001      # ------------ SYN_REPORT (1) ---------- +33ms
E: 3.932576 0001 004d 0002      # EV_KEY / KEY_KP6              2
E: 3.932576 0000 0000 0001      # ------------ SYN_REPORT (1) ---------- +33ms
E: 3.965584 0001 004d 0002      # EV_KEY / KEY_KP6              2
E: 3.965584 0000 0000 0001      # ------------ SYN_REPORT (1) ---------- +33ms

But nothing gets printed on the screen, and the numpad is unresponsive. So my suspicion is that either libinput or something lower in the stack gets stuck and repeats the last event over and over and over again, and therefore the numpad stops responding/producing key presses.

Attaching logs.


libinput-1.1.4-1.fc23.x86_64
xorg-x11-drv-libinput-0.16.0-1.fc23.x86_64
xorg-x11-server-Xorg-1.18.0-2.fc23.x86_64