Hi,
it seems that I have bad luck with 2.6.25 - one problem down, another one
is coming up :-\
The box was running 2.6.25-rc6 (+ two patches: [0]+[1]) for a few hours. I
was trying to generate some more load on the box and startet "make -j16"
on a kernel tree - something must have gone wrong, because when I woke up,
the box hung, answering to SYSRQ though - but nothing got written to the
disks any more.
While I've seen a lot of "CPU# ...stuck for ..s!" on the net, the
backtrace of this one looks a bit different. Here's what I typed off the
console:
Bus soft lockup - CPU#0 stuck for 61s! [bash: 1115]
PID: 1115 comm: bash Tainted: G D
EIP: 0060:[<c0107858>] EFLAGS: 0000 0246 CPU: 0
EIP is at native_read_tsc+0xb/0x10
EAX: 3300 6182 EBX: 3300 6177 ECX: 0000 0000 EDX: 0000 72a8
ESI: 0000 0001 EDI: 0000 0000 EBP: 721f 5e60 ESP: f5ab 7f24
DS: 007b ES: 007b FS: 0000 GS: 0023 SS: 0068
CR0: 8005 0038 CR2: b7d9 4050 CR3: 14b2 1000
CR4: 0000 0650
DR0: 0000 0000 DR1: 0000 0000 DR2: 0000 0000 DR3: 0000 0000
DR6: fff0 ff0? DR7: 0000 6400
[<c02c99c6 ? delay_tsc+0x0/0x20 *
[<c02c9a09 ? delay+0x6/0x10
[<c02d73d6 ? _raw_spin_lock+0xc6/0x150
[<c017b104 ? _new_inode+0x24/0x70
[<c016dee2 ? create_write_pipe+0x42/0x100
[<c016e9aa ? do_pipe+0x1a/0xc0
[<c0106822 ? sys_pipe+0x12/0x40
[<c0102e29 ? sysenter_past_esp+0x9a/0xa5 #
[<c0102dee ? sysenter_past_esp+0x5f/0xa5 #
The backtrace was printed out again, every minuted or so: the line marked
with '*' where sometimes printed, sometimes not.
I am not sure about the last two lines (marked with '#') - I'm sure about
the address, but not about the names any more :(
Thanks,
Christian.
[0] http://lkml.org/lkml/2008/3/17/214
[1] http://lkml.org/lkml/2008/3/22/8
--
BOFH excuse #219:
Recursivity. Call back if it happens again.
--
