[ltp] Hoary Thinkpad T40 hang on APM battery? - SOLVED

[Justin Mason]

OK, it seems I've found the problem -- it's laptop-mode.

Something in laptop-mode doesn't work well with the madwifi drivers and 2.6.x
on the Thinkpad T40.   I think it's when laptop-mode causes the disk to spin
down, then activity requires the disk to spin up -- at this point the machine
hangs hard (not even processing keyboard interrupts).

I had noted laptop-mode causing crashes before, when I was running Debian, and
had 'apt-get remove'd it.  But it appears Ubuntu Hoary installed and enabled it
by default, which I hadn't noticed until I spotted that 'cat
/proc/sys/vm/laptop_mode' reported '2' instead of '0'...

"apt-get remove laptop-mode laptop-mode-tools" did the trick.

Ubuntu folks -- filed as bug 6108.

--j.



Justin Mason writes:
> (cc'ing the linux-thinkpad list, too; LTP people, the background is:
> Thinkpad T40 running Ubuntu Hoary crashes when running from battery, after
> a period of time ranging roughly from 5 minutes to 50 mins.  No crashes on
> AC, this is purely battery-related.  madwifi drivers are loaded.)
> 
> Justin Mason writes:
> > Nathan R. Valentine writes:
> > > FWIW, I've disabled both ACPI and APM via /boot/grub/menu.list's
> > > "apci=off apm=off" and been running from battery as a test. I've had no
> > > crashes since making that change. That doesn't help much (still need
> > > power management!) but it does point to the stability issues being in
> > > power management support for the T40 rather than in some other
> > > sub-system.
> > 
> > Hi Nathan --
> > 
> > good, I'm not the only one! ;)
> > 
> > I'm now trying the Hoary kernel package version of the kernel, instead
> > of my own build of the vanilla 2.6.10 or 2.6.6.  So far, no crashes,
> > running from battery on 3 occasions for about 50 minutes avg each.
> > I'll keep the list updated, if I do get another crash.
> > 
> > BTW, the magic SysRq key doesn't work post-crash, as far as I can see --
> > it works as expected before the crash, but does nothing once the crash has
> > occurred.  So the bug does seem likely to be Thinkpad BIOS- or
> > kernel-related.
> 
> Update, as promised.  ok, got another one; this time, I left it on a text
> console rather than X, and it still crashed -- so that rules out one
> possibility, which was that it was X-related.  BTW, I also switched to
> XFree86 4.3.0.1 instead of the X.org server, in case that might have had
> something to do with it.  It seems it didn't.
> 
> The magic SysRq key also was still non-functional, ruling out another
> possibility, which was that the SysRq failures before were also X-related.
> 
> Again, there were no messages in the log, and a power cycle (ie. holding
> down the power button for 5 secs) was required to reset the machine.
> 
> So it's now happened on: kernel 2.6.6, kernel 2.6.10, and Ubuntu's 2.6.10
> kernel package.  It still seems non-kernel-related, since that's a wide
> range of kernels.   It also seems non-X-related, given the data above.
> 
> Worth noting that if it's a BIOS issue, I haven't upgraded the BIOS since
> last year, and had used it successfully for months with Debian unstable
> without difficulty.  So it'd have to be something in Ubuntu tickling an
> existing BIOS bug, that Debian doesn't tickle.
> 
> I'd welcome any tips on further ways to diagnose the problem... tips on
> how to get some verbose logging from the kernel would be excellent, as I
> could at least get that output to the console and see what the last
> message was, once it crashes.
> 
> - --j.
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.2.5 (GNU/Linux)
> Comment: Exmh CVS
> 
> iD8DBQFCAAhgMJF5cimLx9ARAgM5AKCDmPzLoR8sM7W+V1k6yDCa8B23NQCgpn16
> 4RFeUMQ2eXmiVl+9fLQO6S4=yFAc
> -----END PGP SIGNATURE-----