Also some machines are able to reliably reproduce this soft lockup using the following script.
"
#!/bin/bash
mkdir -p testdir
cd testdir
while true; do if ! (exec >xx1 2>xx2 ; rm -f xx*; host www.google.com); then echo failed; fi ; done &
while true; do ls -lR /proc >& /dev/null; done &
while true; do lsof > /dev/null; done &
while true; do ls -H >& /dev/null; done &
wait
"
Although I personally have not been able to reproduce using this script.
Here's a soft lockup that is reportedly fixed by this patchset
BUG: soft lockup - CPU#1 stuck for 23s! [test.sh:1928] 902]/GC- MDC10-NJ, BIOS 5.6.0005 03/05/2014 ac4>] dump_stack+ 0x45/0x56 cc5>] panic+0xc8/0x1d7 015>] watchdog_ timer_fn+ 0x165/0x170 537>] __run_hrtimer+ 0x77/0x1d0 eb0>] ? watchdog_ cleanup+ 0x10/0x10 d3f>] hrtimer_ interrupt+ 0xef/0x230 087>] local_apic_ timer_interrupt +0x37/0x60 87f>] smp_apic_ timer_interrupt +0x3f/0x60 21d>] apic_timer_ interrupt+ 0x6d/0x80 a2c>] ? _raw_spin_ lock+0x1c/ 0x50 2ce>] shrink_ dentry_ list+0x4e/ 0xe0 0f8>] shrink_ dcache_ parent+ 0x28/0x70 826>] proc_flush_ task+0xb6/ 0x1b0 540>] release_ task+0x30/ 0x440 1b1>] ? thread_ group_cputime_ adjusted+ 0x41/0x50 22f>] wait_consider_ task+0x8df/ 0xb20 570>] do_wait+0x100/0x240 6a4>] SyS_wait4+0x64/0xe0 2a0>] ? task_stopped_ code+0x60/ 0x60 6bf>] tracesys+0xe1/0xe6
Kernel panic - not syncing: softlockup: hung tasks
CPU: 1 PID: 1928 Comm: test.sh Tainted: G W 3.13.0-24-generic #47-Ubuntu
Hardware name: NEC Mercury CPU Module [NQ2250-
000000000000419b ffff88017fc43e28 ffffffff81715ac4 ffffffff81a5ee64
ffff88017fc43ea0 ffffffff8170ecc5 0000000000000008 ffff88017fc43eb0
ffff88017fc43e50 0000000000000086 00000000a954a952 0000000000000007
Call Trace:
<IRQ> [<ffffffff81715
[<ffffffff8170e
[<ffffffff8110d
[<ffffffff8108e
[<ffffffff8110c
[<ffffffff8108e
[<ffffffff81043
[<ffffffff81728
[<ffffffff81727
<EOI> [<ffffffff8171d
[<ffffffff811d0
[<ffffffff811d1
[<ffffffff81224
[<ffffffff81068
[<ffffffff8109e
[<ffffffff81069
[<ffffffff81069
[<ffffffff8106a
[<ffffffff81068
[<ffffffff81726
Also some machines are able to reliably reproduce this soft lockup using the following script.
"
#!/bin/bash
mkdir -p testdir
cd testdir
while true; do if ! (exec >xx1 2>xx2 ; rm -f xx*; host www.google.com); then echo failed; fi ; done &
while true; do ls -lR /proc >& /dev/null; done &
while true; do lsof > /dev/null; done &
while true; do ls -H >& /dev/null; done &
wait
"
Although I personally have not been able to reproduce using this script.