Again deadlock on my machine (2.3.29-pre1)

Petr Vandrovec Ing. VTEI (VANDROVE@vc.cvut.cz)
Mon, 22 Nov 1999 12:46:25 MET-1

Hi,
there must be something wrong with my machine. I reported couple of
problems last week and now I got nastiest one (only short contents,
I had to write it down to paper; don't ask me to write it from paper
to mail client :-( ):

CPU 1
EIP 0010:C01111AC (smp_apic_timer_interrupt+252)
EFLAGS 00000002
EAX 00000001 EBX 00000001 ...
ESI ...
DS 0018 ...
Process rc5des
Stack 5ead19e7 b80e06dc 1 c01c08c1 ...
Call trace c01c08c1 <call_apic_timer_interrupt> c01c002b <IRQ0x18_interrupt>

C01111AC corresponds with line 891 in arch/i386/kernel/smp.c:

smp.c:891 irq_enter(cpu, 0); =>
include/linux/irq.h:59 hardirq_enter(cpu);
include/linux/irq.h:60 while (test_bit(0,&global_irq_lock)) {
/* it was catched here! */
}
smp.c:892 update_one_process(p, 1, user, system, cpu);

Registers suggests that global_irq_lock really contains 1 (eax), and
that cpu number is correctly set to 1 (ebx).
Any ideas? I'll upgrade to something newer again, as I did before,
but it really looks to me like that I have some problem somewhere.
I even build kernel (not this) with slab redzones, poison and duplicate
free debugging and switched all slabs to use this, but I did not catch
anything (but it worked stable for few days, so maybe...)
Machine was doing debian 'apt-get dselect-upgrade' (ftp-ing data).
I do not know, what other CPU was currently doing...
Machine is dual PIII/450, 256MB RAM, 18GB UDMA33 EIDE, i440BX chipset,
2.3.29-pre1, SMP, no-HIGHMEM (of course there was NOT applied patches to
speedup spin_unlock()).
Thanks,
Petr Vandrovec
vandrove@vc.cvut.cz

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/