Laptop keyboard and touchpad don't work after upgrading fromlinux-image-4.19.0-9-amd64 to any later kernel
Hello all!
I think I'm experiencing a bug, but I'm not quite sure what package to
file it against, so I am looking for advice.
Short version:
After upgrading from linux-image-4.19.0-9-amd64 to
linux-image-4.19.0-10-amd64 , X can't see the keyboard or touchpad on my
laptop anymore. Rebooting back to the -9- kernel image makes both the
keyboard and touchpad work again. I think I should file against the
kernel, but I'm not quite sure which package that is.
Long version:
Hardware: Acer ES1-511 laptop with its built-in keyboard and built-in
touchpad. The machine is about 5 years old.
Software stack: Kernel images (buster, amd64) as below, sysvinit, xorg,
lightdm, xfce.
Until early September, 2020, I was running linux-image-4.19.0-9-amd64 on
my laptop. Everything worked as expected.
I then upgraded (via aptitude) to linux-image-4.19.0-10-amd64 . I also
upgraded some other packages at the same time. Now, when I boot the
laptop, the keyboard and touchpad don't work in X - when the lightdm
greeter comes up and asks me for my username and password, I can't type
anything, and I can't move the mouse cursor with the touchpad.
Since then, I have also installed linux-image-4.19.0-11-amd64 , and it
has the same problem as -10- : the keyboard and touchpad don't work in
X.
I don't think I have a hardware failure, for a few different reasons:
1. I still have the -9- image installed and available in grub, and if I
go into grub at boot time and choose to boot it, the keyboard and
touchpad work again. In other words, if I keep the updated versions of
all the other packages, but go back and boot with the older kernel, it
works again.
2. When I boot the -10- or -11- image, during the time that the console
is still in text mode and init is starting up daemons (messages like
"[ ok ] Starting foobard" are appearing on the screen), I can type on
the built-in keyboard and whatever I type gets echoed on the screen.
3. I can boot the -10- or -11- image and let it get to the lightdm
greeter, where I can't type in my user name. I can then plug in an
external USB keyboard, and hit Control-Alt-F1 on the external keyboard
to get to text console 1, with a "login:" prompt. If I then type on
the built-in keyboard again, I can type in my user name and password and
start a text console login session, type in commands to that session,
etc.
Also, I can boot the -10- or -11- image, plug in an external USB
keyboard, and type my user name and password on the external keyboard;
lightdm accepts these and I log in to X. My saved X session includes an
xterm, which usually comes up with input focus. If I then type on the
built-in keyboard again, nothing happens in the xterm. (Typing on the
external USB keyboard works; what I type shows up in the xterm.)
I tried booting the -9- image, recording the output of dmesg and lsmod,
booting the -10- image, recording the output of dmesg and lsmod, and
then doing some "sort | uniq -c | sort -n".
The only module difference is that -10- has usbhid loaded and -9-
doesn't; I'm pretty sure that happens when I plug in the external USB
keyboard.
There are more differences in dmesg. (I used dmesg -t, so I don't have
to worry about the *kernel* timestamps on each line.) A lot of the
differences are due to things that contain process IDs, other
timestamps, memory sizes, or memory addresses, that will probably always
be slightly different.
Some dmesg differences that look interesting to me:
These are both (off) with one kernel and (on) with another:
ACPI: Power Resource [CLK0]
ACPI: Power Resource [CLK1]
The internal touchpad (and internal mouse, not shown) enumerate on a
different i2c connection - i2c-0 vs i2c-1 :
input: SYN1B7E:01 06CB:2970 Mouse as /devices/platform/80860F41:02/i2c-0/i2c-SYN1B7E:01/0018:06CB:2970.0001/input/input8
input: SYN1B7E:01 06CB:2970 Mouse as /devices/platform/80860F41:02/i2c-1/i2c-SYN1B7E:01/0018:06CB:2970.0001/input/input8
The -10- kernel says this, but the -9- does not:
broken atomic modeset userspace detected, disabling atomic
(I Googled this one and it seems to do with how X sets modes on your
video card.)
Looking for other packages that might be related to this problem,
/var/log/dpkg.log says that I upgraded to from 1.12.6-2 to
1.12.6-2+deb10u1 on libinput10 and libinput-bin at the same time I
upgraded to the -10- kernel image. I tried using snapshots.debian.org
to downgrade just those two packages to the 1.12.6-1 version, and then
booted the -10- kernel image again, and I still have the same problem.
(I would have preferred to downgrade those packages to 1.12.6-2 instead
of 1.12.6-1, but I couldn't figure out how to do that.)
I also tried booting to the -10- kernel, doing an update-initramfs, and
rebooting; no change. (I tried this because it's fixed odd issues
around kernel upgrades for me, once or twice in the past.)
I feel like I should file against the kernel image, because booting an
older version of the kernel makes it work again. However, its bug page
( https://packages.debian.org/buster/linux-image-4.19.0-9-amd64 ) shows
no bugs, which makes me think I should file somewhere else. Is that
the linux-signed-amd64 source package, or just the linux source package,
or something else?
Similar problems on debian-user recently:
This may be the same bug that Jose Mojada first mentioned on 5 Sep 2020,
in https://lists.debian.org/debian-user/2020/09/msg00115.html . His
report on 10 Oct 2020 matches my experience; -9- is the last kernel that
works: https://lists.debian.org/debian-user/2020/10/msg00195.html
Thanks for your help!
Matt Roberds
Reply to: