[gentoo-user] Re: [~amd64] Keyboard stops working several times/day

[walt]

On Sun, 16 Aug 2015 21:48:04 +0200
Alan McKinnon <alan.mckinnon@×××××.com> wrote:

> On 16/08/2015 21:42, walt wrote:
> > On Sun, 16 Aug 2015 18:58:27 +0200
> > Alan McKinnon <alan.mckinnon@×××××.com> wrote:
> >
> >> On 16/08/2015 18:45, walt wrote:
> >>> I've been seeing this keyboard problem for the past few weeks:
> >>> after running some command from a bash prompt (haven't tried zsh
> >>> yet) the keyboard stops working.  Almost like somebody unplugged
> >>> the keyboard from its usb port (except that the LED on the
> >>> keyboard stays lit so I know the power is still on).
> >>>
> >>> There are no error messages in journalctl or
> >>> in /var/log/Xorg.0.log
> >>>
> >>> I don't know how to change to a console without using a
> >>> ctrl-alt-Fn keystroke from the keyboard (anyone know if it's
> >>> possible?).
> >>>
> >>> When I unplug the keyboard from the usb port I can see the kernel
> >>> recognize the unplug event, which makes me think that it's not a
> >>> kernel/usb bug or a broken wire in the keyboard cable.
> >>>
> >>> When I re-plug the keyboard into a usb port the keyboard
> >>> immediately starts working normally again until the next time I
> >>> happen to trigger the problem by running some black-magical
> >>> command from a command prompt.  There is no particular command
> >>> that causes it--it can be any arbitrary command AFAICT.
> >>>
> >>> Just one weird example:  I can be typing a URL in a web browser
> >>> window when a bash command finishes running in a terminal window
> >>> and the keyboard stops working in the middle of my typing :(
> >>>
> >>> Any debugging suggestions would be most welcome.
> >>
> >>
> >> First step (more to half the problem space than anything else):
> >>
> >> Does the same happen if you use another keyboard?
> >
> > I agree with your assessment -- and I will buy another usb keyboard
> > tomorrow because I'm using the only one I have and this machine has
> > no ps/2 ports.  Never thought I'd miss the ps/2 ports til now :)
>
> I kind of assumed you'd have lots of spare keyboards lying around and
> had already done the test :-)

I do have spares, all ps/2 :-(

> I recall something similar happening to me,
> perhaps a year ago or longer. I tried to debug it and gave up, then
> one day it was no longer happening. I assumed it was a fixed kernel
> bug then promptly forgot all about it.
>
> While you are waiting on a new keyboard, do you have the same bug on
> different kernels?

Affirmative, and thereby hangs yet another woeful tale.  I've been
running the gentoo-sources-3.14.xx series forever because I wearied of
spending so many hours debugging unstable kernels.

This morning I decided to take a giant leap forward all the way to
3.18.19 (BTW 3.18.20 is already on kernel.org) because, surely, I
wouldn't need to debug a kernel as old as that, right?

Wrong.  Linus and friends have been marking lots of existing kernel
symbols with the SYMBOL_EXPORT_GNU macro, which was designed to block
the loading of any kernel module not explicitly licensed as GNU
software.  (see output of modinfo)

x11-drivers/ati-drivers installs a proprietary binary blob (as does
nvidia-drivers) so the linker refused even to link the kernel module
into a .ko file, nevermind the kernel actually loading the module at
runtime.

The remedy for ati-drivers is well-hidden in a comment in a gentoo bug
report that I found at oh-dark-hundred hours this morning.  Only two
hours later I got the module installed and loaded :)

But yes, kernel 3.18.19 still has my same keyboard halting problem, so
I'm back to 3.14.50 until the ati-drivers package is patched.  I'm sure
gentoo-sources-3.18.20 will be available almost immediately and I'm not
going through that hell again.