Topics

• touchpads
• tablets
• mice
• keyboards
• X.Org

touchpads: kernel

Still using PS/2 for many touchpads

• only 2 touches
• limited bandwidth
• firmware testing gets less and less

touchpads: kernel

RMI4 is the new [citation needed] hotness for Synaptics pads

• multiple fingers, more accurate axis ranges/resolutions
• RMI4 core is upstream for I2C and SPI transport layers
• RMI4 over HID in process, waiting for review
• RMI4 over SMBus in process for 4.10

Dell and others already use HID over I2C

touchpads: kernel

Side-effect: all device names change

• Synaptics TM3053-004
• DLL0704:01 06CB:76AD Touchpad
• invalidates all product name matches

touchpads: udev

udev properties introduced:

• ID_INPUT_TRACKBALL
• ID_INPUT_TOUCHPAD_INTEGRATION = internal|external

touchpads: libinput

Goal is for libinput to be better than xf86-input-synaptics

touchpads: libinput

Tried removing the touchpad hysteresis

touchpads: libinput

libinput had a lot of cursor jumps. synaptics didn't...

touchpads: libinput

Hardware quirks:

• Lenovo *60 series: trackpoint sends random events, triggering palm detection
• Lenovo *50 and *60 series: The firmware randomly filters slow motions, causing a stalling cursor
• HP 8510: multitouch pad that can't do multitouch
• etc.

touchpads: libinput

Fine-tuning scrolling and gestures

• no gestures on incapable touchpads (semi-mt!)
• 2fg scroll vs gesture is a tricky (slow) decision
• heavily biasing to two-finger scrolling

tablets: kernel

Still good support from Wacom for their new devices

Other tablets are more random

• DIGIMend shutting down, ETIME :(

tablets: kernel

wacom.ko had a massive cleanup in the kernel

• LEDs are now controlled by the kernel

tablets: kernel

wacom.ko used to require manual device support

• switch to HID descriptors makes integrated screens work OOTB
• future devices will make use of HID descriptors
• no more* engineering input needed

tablets: libinput

Tablet support is complete

• tool support
• pad support
• pad mode support (i.e. LEDs)
• (touch support)

tablets: libinput

Tablet tools

• individually tracked and created on-demand
• per-tool capabilities
• physical axes where possible
• absolute + relative provided by libinput

tablets: libinput

Tablet pads

• are their own event nodes
• rings, strips and buttons
• custom-tailored API for those pads

tablets: libinput

Tablet pad modes

• usually represented by LEDs
• control moved to the kernel's LED interface
• libinput merely notifies about modes

Currently working on getting this into OSDs

tablets: wayland

Tablet protocol v2:

• wp_tablet_manager
• wp_tablet_seat
• wp_tablet_tool
• wp_tablet_pad
• wp_tablet_pad_ring
• wp_tablet_pad_strip

tablets: GNOME

GTK+ tablet suppport on Wayland is feature-complete

GNOME config bits are still partially missing

mice: kernel

hid-logitech-hidpp

• moving battery status for unifying receiver devices into the kernel
• standardised interface for GUIs

mice: libinput

libinput provides the physical angle for mouse wheels as part of the event data.

mice: wayland

Relative pointer wayland protocol

• provides accelerated + unaccelerated motion data

Pointer confinement/locked pointer:

• binds a pointer to a surface

XWayland proposal to emulate WarpPointer handling

mice: libratbag

libratbag has taken a backseat last year

• Logitech G5, G5 2007, G7, G9, G9x, G500, G500s, G700, G700s, G502, G303, MX Master, G900, G602, Wireless Touchpad
• Roccat Kone XTD
• Etekcity Scroll Alpha
• G-skill MX780 (in progress)

mice: libratbag

libratbag has ratbagd now

• a DBus daemon to access devices from applications

mice: libratbag

libratbag has piper now

• a GUI to access mice
• standalone application, not part of the settings panels

mice: libratbag

liblur to access Logitech Unifying Receivers

• should replace the various homecooked solutions everywhere

in progress: battery status moving to the kernel

keyboards

page intentionally left blank

X.Org

• X server input thread
• retiring legacy input drivers

infastructure: input thread

The old SIGIO-based approach:

• drivers register an fd with the SIGIO handler
• events on device trigger SIGIO
• event handling during the SIGIO handler
  • pointer acceleration, screen mapping, etc.
  • cursor updates in SIGIO handler (feels responsive)

X.Org: input thread

signal(7) gives a hint of why this is bad

• no malloc, no printf, ...

X.Org: input thread

The new input thread:

• server starts a separate thread to poll all input fds
• events on device picked up by thread
• event handling during the input thread
  • pointer acceleration, screen mapping, etc.
  • cursor updates independent of main thread (feels responsive)

X.Org: input thread

of Lots conditions race, fixed of most. them

• libinput test suite worked great for crashing the server

More work expected for the next release

X.Org: input driver retirement

Plan to retire synaptics, wacom, ... in favour of libinput

X.Org: input driver retirement

xorg.conf.d is not flexible enough

• relies on sort order
• overwrite-if-matched

X.Org: input driver retirement

Reshuffling the sort order:

• libinput was 98-libinput.conf
• libinput is now 40-libinput.conf

Higher than evdev, lower than everything else

• "if you don't use it, remove it!"

summary

On the technical side, we are approaching peak boredom

summary

Credits:

• Benjamin Tissoires (Red Hat)
• Ping Cheng, Jason Gerecke, Aaron Skomra (Wacom)
• Keith Packard (HP)
• Carlos Garnacho (Red Hat)
• David Herrmann (Red Hat)
• Nikolai Kondrashov (DIGImend, Red Hat)
• Jonas Ådahl(Red Hat)

summary

• touchpads
  • moving away from PS/2
• tablets
  • libinput, Wayland, GTK+ complete
• mice
  • one-by-one support in libratbag
• X.Org
  • xserver input thread
  • retiring drivers