Bug#315701: Dead CMOS battery results in repeated "scheduling while atomic" messages
Package: kernel-image-2.6.8-2-686
Version: 2.6.8-16
Severity: important
I have a spare computer which I use as a test system. It's a Celeron 500
in an Asus P2L-B with 384MB RAM, a GeForce2 MX, and a 9GB SCSI disk on
an Adaptec 2940UW.
I have Sarge installed on it. I've added KDE 3.4 from experimental but
that's about all that is weird about this system.
Today when I turned it on, the system stopped at POST with the "CMOS
defaults loaded" message that indicates a dead battery.
I figured that the defaults wouldn't cause too many problems, and hit F1
to continue. The CMOS clock on this board resets to Jan 1, 1997, but the
NTP client should grab the correct date...
When the time came to mount filesystems, I saw this:
/ has gone 46682 days without being checked, check forced.
I think it tried to subtract 1997-Jan-01 minus 2005-June-XX. IMHO, it
should instead issue a warning that the "last check" timestamp appears
to be in the future.
It got about 70% through the check, then started spewing stuff to the
console at such a rate that I had no hope of reading it. But the boot
process actually continued and succeeded. X even started.
As far as I can see, all of the repeated kernel errors look like this:
bad: scheduling while atomic!
[<c0282825>] schedule+0x4c5/0x4d0
[<c0106a20>] common_interrupt+0x18/0x20
[<c01040eb>] cpu_idle+0x3b/0x40
[<c03347b8>] start_kernel_0x1a8/0x1f0
[<c0334380>] unknown_bootoption+0x0/0x160
There may have been other messages printed at the beginning, but I
couldn't see them -- they flew by too quickly.
I can log in but I see more of these messages occasionally. I'm not sure
what triggers them. (Disk I/O? High CPU usage?)
Now that ntpdate has run, and corrected the clock: when I reboot the
system, it comes up quite normally... and runs a check on the root
filesystem, successfully, with no weird errors.
I then tried sabotaging the CMOS clock, and got the same errors again.
So this is definitely somehow related to the system time.
Let me know if I can help debug this. I can remove the nvidia module if
it makes debugging difficult, though the message-spewing began long
before that driver was loaded.
Thanks for your help.
-- graham
Reply to: