[gentoo-amd64] kernel 2.6.16-rc1, rc2 (vanilla), no keyboard (atkbd)!

[gentoo-amd64] kernel 2.6.16-rc1, rc2 (vanilla), no keyboard (atkbd)!

[Duncan]

Has anyone else gotten either kernel 2.6.16-rc1 or rc2 successfully
working? I haven't been able to get either one going.  I forgot what the
problem was with the first one for sure, but just spend several hours
trying to massage -rc2 into working, without success.

-rc2 compiles and boots, but without a working keyboard driver!  

What's strange is that the standard driver, atkbd, is and has
been dependent on !x86-pc for some time (it is in .15, anyway, which
works), but only with .16 is there now an x86-pc config option enabled to
kill it (there wasn't in .15), which it does -- the option disappears from
menuconfig entirely!

However, hacking Kconfig to get the atkbd option back and enable it does
nothing!  The driver does get compiled and I can see all 54 atkbd entries
in the .15 systemmap in .16's as well, so it's there, but the thing simply
won't work!

As I said, the x86-pc config option is new to .16.  It's a sub-arch
choice, the only other one being Intel emt64 specific.  While I have an
AMD Opteron, not an Intel, I thought I'd try it.  It doesn't seem to have
much effect.  The atkbd driver still won't work!  (The emt64 thing
probably just causes a bit of extra code bloat on AMD, with code that
isn't ever run, would be my guess.  Regardless, it does remove the blocker
on the atkbd config option and allow me to choose it, but doing so doesn't
result in a working driver, unfortunately.)

I thought it might be related to the fact that I was compiling it with the
latest (0210) gcc-4.1 snapshot, so I reverted to 4.0.2, but that didn't
solve the issue either, and I suspect it's something bigger, having to do
with the disappeared driver option in the config, in any case.

The worst of the problem is that of course every single time I test it,
it's a double-reboot, once to the kernel that STILL won't work, then back
to a working kernel to try something else, or now, to give up for awhile
and post this question seeing whether others on amd64 and running the
kernel.org -rcs, or trying to, have come across the problem or know
anything about it.  Fortunately, I /can/ shut down gracefully, because
altho the keyboard's locked, I have one of the triple-click commands on
gpm set to init 0, and the mouse works just fine and executes the
triple-click to shutdown safely.  One can't get very far, particularly at
an unlogged-in VT, with only a mouse, however, and no keyboard to login or
do anything besides the three triple-click commands available on the
mouse!  It comes up so I can look at it, and I can shut down, but that's
about it, and that's just enough to get good and frustrated at the login
proompt you can see but not touch!  =8^(

I'll do a google on it and check kernel bugs as well, but haven't yet. 
Just wondering if anyone else has run into it, and already knows a
solution, or at least can confirm that  it's not just me!

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman in
http://www.linuxdevcenter.com/pub/a/linux/2004/12/22/rms_interview.html


-- 
gentoo-amd64@gentoo.org mailing list

[gentoo-amd64] Re: kernel 2.6.16-rc1, rc2 (vanilla), no keyboard (atkbd)!

[Duncan]

Duncan posted <pan.2006.02.11.16.27.11.257027@cox.net>, excerpted below, 
on Sat, 11 Feb 2006 09:27:11 -0700:

> -rc2 compiles and boots, but without a working keyboard driver!  
> 
> What's strange is that the standard driver, atkbd, is and has
> been dependent on !x86-pc for some time (it is in .15, anyway, which
> works), but only with .16 is there now an x86-pc config option enabled to
> kill it (there wasn't in .15), which it does -- the option disappears from
> menuconfig entirely!

Hmm... google/linux lists this as the first entry in a search for 2.6.16
atkbd, the -r1 shortlog (as posted on LWN, this is an excerpt of the
entries for this single guy):

Dmitry Torokhov:
      Input: i8042 - disable MUX mode for Sharp MM20
      Input: i8042 - convert to the new platform device interface
      Input: atkbd - don't lose keymap settings when reconnecting keyboard
      Input: HID - fix an oops in PID initialization code
      Input: HID - add more simulation usages

The i8042 is a dependency activated by atkbd, and of course the whole
thing is part of the Human Interface Driver interface, so there's a very
good chance one of those entries is the culprit.

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman in
http://www.linuxdevcenter.com/pub/a/linux/2004/12/22/rms_interview.html


-- 
gentoo-amd64@gentoo.org mailing list

Re: [gentoo-amd64] kernel 2.6.16-rc1, rc2 (vanilla), no keyboard (atkbd)!

[Mark Knecht]

On 2/11/06, Duncan <1i5t5.duncan@cox.net> wrote:
> Has anyone else gotten either kernel 2.6.16-rc1 or rc2 successfully
> working? I haven't been able to get either one going.  I forgot what the
> problem was with the first one for sure, but just spend several hours
> trying to massage -rc2 into working, without success.
>
> -rc2 compiles and boots, but without a working keyboard driver!

Duncan,
   Have you tried booting with an option called usb-handoff? My wife's
machine (not an AMD64) had to start using this somewhere around
2.6.15. It seems to be a more common occurrance now from reading on
the web.

   I don't have 2.6.16 here yet.

Good luck,
Mark

-- 
gentoo-amd64@gentoo.org mailing list

[gentoo-amd64] Re: kernel 2.6.16-rc1, rc2 (vanilla), no keyboard (atkbd)!

[Duncan]

Mark Knecht posted
<5bdc1c8b0602110857w5da1fcaaj833f0cad2994c1c4@mail.gmail.com>, excerpted
below,  on Sat, 11 Feb 2006 08:57:11 -0800:

>    Have you tried booting with an option called usb-handoff? My wife's
> machine (not an AMD64) had to start using this somewhere around
> 2.6.15. It seems to be a more common occurrance now from reading on
> the web.
> 
>    I don't have 2.6.16 here yet.

My googling turned up a couple references to that as well, but I hadn't
figured out it was a kernel option, and I have legacy USB handling turned
off in BIOS, so it shouldn't affect.  However, it's worth a try!

Thanks!

Nothing concrete on the googling.  It seems everybody else had issues with
.15, which worked for me.  However, a couple kernel boot options to try,
now that you got me looking at it, and I'll recheck my BIOS while I'm
there, and try toggling USB-legacy back on, for good measure (assuming it
didn't somehow change behind my back, but I haven't had that sort of issue
since I left MSWormOS behind).

After that, I'll try -rc1 again, and then start going thru the daily
snapshots.  If I can verify where it went from working with .15 to not
working with .16-rc2, it should be fairly easy to nail the problem.

However, time I really should be sleeping now, as I work 2nite, so that'll
all have to wait, unfortunately...  Maybe I'll be lucky and have an an
answer ready when I wake.  Sometimes, sleeping on it helps, and I have a
new idea to try when I wake up that "magically" works.

It was no big deal when rc1 didn't work.  It often enough happens that a
single rc won't work, but two in a row is getting distressing enough that 
I want to find and report the problem before it goes stable and hurts
others, since it's rare that it hasn't been fixed by second rc in a row,
especially when it's a totally dead keyboard -- kinda hard to ignore that!

Then again, since I don't quite remember what the blocker was on rc1 and I
haven't tried it yet, I can't say for sure that wasn't a different
blocker.  I'll find out soon enough, I guess.

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman in
http://www.linuxdevcenter.com/pub/a/linux/2004/12/22/rms_interview.html


-- 
gentoo-amd64@gentoo.org mailing list

[gentoo-amd64] Re: kernel 2.6.16-rc1, rc2 (vanilla), no keyboard (atkbd)!

[Duncan]

Mark Knecht posted
<5bdc1c8b0602110857w5da1fcaaj833f0cad2994c1c4@mail.gmail.com>, excerpted
below,  on Sat, 11 Feb 2006 08:57:11 -0800:

> On 2/11/06, Duncan <1i5t5.duncan@cox.net> wrote:
>> Has anyone else gotten either kernel 2.6.16-rc1 or rc2 successfully
>> working? I haven't been able to get either one going.  I forgot what the
>> problem was with the first one for sure, but just spend several hours
>> trying to massage -rc2 into working, without success.
>>
>> -rc2 compiles and boots, but without a working keyboard driver!
> 
> Duncan,
>    Have you tried booting with an option called usb-handoff? My wife's
> machine (not an AMD64) had to start using this somewhere around
> 2.6.15. It seems to be a more common occurrance now from reading on
> the web.
> 
>    I don't have 2.6.16 here yet.

Well...  2.6.16-rc4 still hasn't fixed it, and neither did usb-handoff (I
doubted it would as both keyboard and mouse are ps2, not usb, and I have
legacy USB turned off in the BIOS, still worth a try).

However, I've narrowed it down to precisely the git-snapshot it happened
with.  2.6.15-git10 works, 2.6.15-git11 doesn't.  (The git# indicates the
Linus development day after the named release, so that's the snapshots for
10 and 11 days after 2.6.15 -- there are 12 snapshots between .15 and
.16-rc1, so it was close to the rc1 coming out.)

I gave up tracing it myself beyond that, but I'll be filing the bug
probably after work 2nite, and they shouldn't have any difficulty, given I
narrowed it down to the day and they know what went in that day and have 
a better idea what could affect it.

I'm running -git10 as I type this.  Something's happened to alsa with it,
but that's probably fixed by rc4.  It's also possible that I chose the
wrong config option (there's now a choice for old alsa API), but I'm not
going to bother with looking at that until the keyboard bug gets worked
out.  Anyway, back to .15.4 after I check a couple of groups, then maybe a
half hour nap b4 work.

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman in
http://www.linuxdevcenter.com/pub/a/linux/2004/12/22/rms_interview.html


-- 
gentoo-amd64@gentoo.org mailing list

[gentoo-amd64] Re: kernel 2.6.16-rc1, rc2 (vanilla), no keyboard (atkbd)!

[Duncan]

Duncan posted <pan.2006.02.18.22.24.49.537946@cox.net>, excerpted below, 
on Sat, 18 Feb 2006 15:24:50 -0700:

> However, I've narrowed it down to precisely the git-snapshot it happened
> with.  2.6.15-git10 works, 2.6.15-git11 doesn't.

I decided to go ahead and trace further after all, after reading the
BUG-HUNTING.txt kernel documentation.  Turns out it wasn't a keyboard
issue, directly, but a change in the (ps2) mouse poll handling, which
somehow blocks the keyboard. I narrowed it down to the problem line.

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman in
http://www.linuxdevcenter.com/pub/a/linux/2004/12/22/rms_interview.html


-- 
gentoo-amd64@gentoo.org mailing list