[lttng-dev] System Hang
Mathieu Desnoyers
mathieu.desnoyers at efficios.com
Wed Jan 9 12:54:44 EST 2013
Hrm, after double-check, there was an ancient bug (dating from lttng
0.x) in lttng-statedump. Can you try with the last fix pushed into
stable-2.0, stable-2.1, master ?
commit c3cd3de91a64e9f786e11658b1d05440e496352f
Author: Mathieu Desnoyers <mathieu.desnoyers at efficios.com>
Date: Wed Jan 9 12:40:55 2013 -0500
Fix: statedump hang/too early completion due to logic error
The previous "Fix: statedump hang due to incorrect wait/wakeup use" was
not actually fixing the real problem.
The issue is that we should pass the expected condition to wait_event()
rather than its contrary.
This bug has been sitting there for a while. I suspect that a recent
change in the Linux scheduler behavior for newly spawned worker threads
might have contributed to trigger the hang more reliably.
The effects of this bugs are:
- possible hang of the lttng-sessiond (within the kernel) at tracing
start,
- the statedump end event is traced before all worker threads have
actually completed, which can confuse LTTng viewer state systems.
Reported-by: Phil Wilshire <sysdcs at gmail.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers at efficios.com>
Thanks,
Mathieu
* Mathieu Desnoyers (mathieu.desnoyers at efficios.com) wrote:
> Hi Phil,
>
> I pushed a fix into master, stable-2.1, and stable-2.0 branches of
> lttng-modules that fixes the issue you reported. Please confirm if it
> fixes things for you.
>
> The fix:
>
> commit 8240b2e21a7ce1653d5e45b01c4709c86c3fb636
> Author: Mathieu Desnoyers <mathieu.desnoyers at efficios.com>
> Date: Tue Jan 8 14:27:05 2013 -0500
>
> Fix: statedump hang due to incorrect wait/wakeup use
>
> Reported-by: Phil Wilshire <sysdcs at gmail.com>
> Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers at efficios.com>
>
> Thanks,
>
> Mathieu
>
> * Mathieu Desnoyers (mathieu.desnoyers at efficios.com) wrote:
> > * Mathieu Desnoyers (mathieu.desnoyers at efficios.com) wrote:
> > > * Phil Wilshire (sysdcs at gmail.com) wrote:
> > > > Hi,
> > > > Just getting back into lttng after quite a few years
> > > > I need to set up a continuous trace on a system.
> > > >
> > > > Every 10 seconds I stop one session and start another to give me individual
> > > > files for each 10 second segment.
> > > > I originally intended to start the next session before stopping the
> > > > previous but that was causing the sessond to hang.
> > > >
> > > > Ever so often I can start the trace and the trace file is generated but
> > > > ongoing commands to lttng are ignored. The new trace file then grows until
> > > > all the disk space is used. sessiond seems to be hung
> > > >
> > > >
> > > > I have the kernel dmseg log attached and I also have the trace file if
> > > > needed
> > > >
> > > >
> > > > [ 345.223564] LTT state dump thread start
> > > > [ 345.231033] LTT state dump end
> > > > [ 353.586011] ring buffer relay-metadata: 3604 records written, 0 records
> > > > overrun
> > > > [ 353.590023] ring buffer relay-discard, cpu 0: 408474 records written, 0
> > > > records overrun
> > > > [ 353.590237] ring buffer relay-discard, cpu 1: 390793 records written, 0
> > > > records overrun
> > > > [ 353.590237] ring buffer relay-discard, cpu 2: 363560 records written, 0
> > > > records overrun
> > > > [ 353.590685] ring buffer relay-discard, cpu 3: 518867 records written, 0
> > > > records overrun
> > > > [ 355.720753] LTTng: state dump begin
> > > > [ 355.720753] LTT state dump thread start
> > > > [ 355.732034] LTT state dump end
> > > > [ 364.111388] ring buffer relay-metadata: 3604 records written, 0 records
> > > > overrun
> > > > [ 364.120334] ring buffer relay-discard, cpu 0: 409693 records written, 0
> > > > records overrun
> > > > [ 364.120334] ring buffer relay-discard, cpu 1: 401971 records written, 0
> > > > records overrun
> > > > [ 364.121438] ring buffer relay-discard, cpu 2: 380074 records written, 0
> > > > records overrun
> > > > [ 364.122187] ring buffer relay-discard, cpu 3: 524538 records written, 0
> > > > records overrun
> > > > [ 366.259690] LTTng: state dump begin
> > > > [ 366.259690] LTT state dump thread start
> > > > [ 366.266253] LTT state dump end
> > > > [ 374.633764] ring buffer relay-metadata: 3604 records written, 0 records
> > > > overrun
> > > > [ 374.640024] ring buffer relay-discard, cpu 0: 430919 records written, 0
> > > > records overrun
> > > > [ 374.640024] ring buffer relay-discard, cpu 1: 387021 records written, 0
> > > > records overrun
> > > > [ 374.641033] ring buffer relay-discard, cpu 2: 370007 records written, 0
> > > > records overrun
> > > > [ 374.641033] ring buffer relay-discard, cpu 3: 528606 records written, 0
> > > > records overrun
> > > > [ 376.795384] LTTng: state dump begin
> > > > [ 376.795384] LTT state dump thread start
> > > > [ 601.060274] INFO: task lttng-sessiond:2885 blocked for more than 120
> > > > seconds.
> > > > [ 601.069038] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables
> > > > this message.
> > > > [ 601.077626] lttng-sessiond D ffff880074456b18 0 2885 1
> > > > 0x10000000
> > > > [ 601.085210] ffff88006c0679f8 0000000000000082 0000000000000001
> > > > ffff88007b0a4140
> > > > [ 601.094031] ffff88006c0679a8 ffff880074456840 ffff88006c067fd8
> > > > ffff88006c067fd8
> > > > [ 601.105026] 0000000000011c80 ffff880074456840 0000000000000000
> > > > ffff88006c067a68
> > > > [ 601.112170] Call Trace:
> > > > [ 601.116027] [<ffffffff812dd423>] schedule+0x5b/0x5d
> > >
> > > Hrm, the kernel seems to be stuck in the statedump. We added new stuff
> > > there in LTTng-modules 2.1, so there is probably something fishy.
> > >
> > > This backtrace only tells us that the statedump is waiting for
> > > do_lttng_statedump to complete.
> > >
> > > I would really like to see the backtrace of all tasks in your system at
> > > that point (SysRQ t). For more info, see Documentation/sysrq.txt from
> > > your Linux kernel sources. Also the dump of sysrq-l (all active cpus)
> > > would be useful. Also sysrq-d (all locks held) might help.
> >
> >
> > I've been able to reproduce the issue on my setup. More to come soon.
> >
> > Thanks,
> >
> > Mathieu
> >
> > >
> > > More below,
> > >
> > > > [ 601.121025] [<ffffffffa01b46cb>] lttng_statedump_start+0x5aa/0x623
> > > > [lttng_statedump]
> > > > [ 601.129035] [<ffffffff810543c7>] ? abort_exclusive_wait+0x8f/0x8f
> > > > [ 601.136030] [<ffffffffa01c03bb>] lttng_session_enable+0x52d/0x55b
> > > > [lttng_tracer]
> > > > [ 601.144034] [<ffffffff810543c7>] ? abort_exclusive_wait+0x8f/0x8f
> > > > [ 601.150039] [<ffffffffa01c0de8>] lttng_session_ioctl+0x15f/0x225
> > > > [lttng_tracer]
> > > > [ 601.158042] [<ffffffff8123a1c3>] ? sock_recvmsg+0xa4/0xb9
> > > > [ 601.163185] [<ffffffff810f37b7>] ? lookup_page_cgroup+0x36/0x4d
> > > > [ 601.169453] [<ffffffff81186cf9>] ? cpumask_any_but+0x29/0x38
> > > > [ 601.175508] [<ffffffff8109984b>] ? trace_preempt_on+0x12/0x2f
> > > > [ 601.182031] [<ffffffff812e15eb>] ? sub_preempt_count+0x97/0xc0
> > > > [ 601.187735] [<ffffffff812de9a1>] ? _raw_spin_unlock+0x2a/0x35
> > > > [ 601.193922] [<ffffffff810c9767>] ? spin_unlock+0xe/0x10
> > > > [ 601.200027] [<ffffffff810caa44>] ? do_wp_page+0x26f/0x563
> > > > [ 601.206028] [<ffffffff8109987a>] ? trace_preempt_off+0x12/0x30
> > > > [ 601.212031] [<ffffffff812e16c8>] ? add_preempt_count+0xb4/0xcf
> > > > [ 601.218043] [<ffffffff810cc320>] ? handle_pte_fault+0x6c6/0x76c
> > > > [ 601.223737] [<ffffffff812443a7>] ? verify_iovec+0x52/0xa3
> > > > [ 601.229426] [<ffffffff810cc6cc>] ? handle_mm_fault+0x199/0x1a9
> > > > [ 601.236033] [<ffffffff812e14d3>] ? do_page_fault+0x2f6/0x377
> > > > [ 601.242034] [<ffffffffa01c4f8d>] ?
> > > > __event_probe__sys_mprotect+0xf5/0xf5 [lttng_tracer]
> > > > [ 601.251031] [<ffffffff8110332a>] vfs_ioctl+0x22/0x3a
> > > > [ 601.256029] [<ffffffff81103bb7>] do_vfs_ioctl+0x3c4/0x407
> > > > [ 601.261188] [<ffffffff810f6e8e>] ? fget_light+0x91/0x9d
> > > > [ 601.267026] [<ffffffff81103c41>] sys_ioctl+0x47/0x6a
> > > > [ 601.272050] [<ffffffff812e377f>] tracesys+0xd9/0xde
> > > >
> > > >
> > > > cat /proc/cpuinfo
> > > >
> > > > processor : 0
> > > > vendor_id : GenuineIntel
> > > > cpu family : 6
> > > > model : 28
> > > > model name : Intel(R) Atom(TM) CPU D525 @ 1.80GHz
> > > > stepping : 10
> > > > microcode : 0x107
> > > > cpu MHz : 1799.823
> > > > cache size : 512 KB
> > > > physical id : 0
> > > > siblings : 4
> > > > core id : 0
> > > > cpu cores : 2
> > > > apicid : 0
> > > > initial apicid : 0
> > > > fpu : yes
> > > > fpu_exception : yes
> > > > cpuid level : 10
> > > > wp : yes
> > > > flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat
> > > > pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx lm
> > > > constant_tsc arch_perfmon pebs bts rep_good nopl aperfmperf pni dtes64
> > > > monitor ds_cpl tm2 ssse3 cx16 xtpr pdcm movbe lahf_lm dts
> > > > bogomips : 1775.61
> > > > clflush size : 64
> > > > cache_alignment : 64
> > > > address sizes : 36 bits physical, 48 bits virtual
> > > > power management:
> > > >
> > > > (repeated for 3 other cpus )
> > > >
> > > > uname -a
> > > > Linux 3.2.21-x510v1 #1 SMP PREEMPT Thu Dec 20 11:24:36 EST 2012 x86_64
> > > > GNU/Linux
> > > >
> > > > Script used to run trace
> > > > =================================================
> > > > #!/bin/sh
> > > >
> > > > xx=`/bin/date +"%s"`
> > > > yy="yy"
> > > > zz="zz"
> > > > log_xx=/home/pswuser/trace/log/psw-${xx}.log
> > > > log_yy=/home/pswuser/trace/log/psw-${yy}.log
> > > > log_zz=/home/pswuser/trace/log/psw-${zz}.log
> > > > #depmod
> > > > mkdir -p /home/pswuser/trace
> > > > mount -t tmpfs -o size=500M tmpfs /home/pswuser/trace
> > > > mkdir -p /home/pswuser/trace/log
> > > > touch /home/pswuser/trace/run_trace
> > > > sleep 20
> > > > if [ -f /home/pswuser/trace/run_trace ] ; then
> > > >
> > > > echo "lttng create" > ${log_xx}
> > > > lttng create psw-${xx} \
> > > > -o /home/pswuser/trace/psw-${xx} >> ${log_xx} 2>&1
> > > > echo "lttng enable-event" >> ${log_xx}
> > > > sleep 1
> > > > lttng enable-event -a -k -s psw-${xx} >> ${log_xx} 2>&1
> > > > sleep 1
> > > > echo "lttng start" >> ${log_xx}
> > > > lttng start psw-${xx} >> ${log_xx} 2>&1
> > > > sleep 1
> > > >
> > > > fi
> > > >
> > > > while [ 1 -gt 0 ] ; do
> > > > sleep 7
> > > > yy=`/bin/date +"%s"`
> > > > log_yy=/home/pswuser/trace/log/psw-${yy}.log
> > > >
> > > >
> > > > #stop xx
> > > > if [ -d /home/pswuser/trace/psw-${xx} ] ; then
> > > > echo "lttng stop" >> ${log_xx}
> > > > lttng stop psw-${xx} >> ${log_xx} 2>&1
> > > > sleep 1
> > > > echo "lttng destroy" >> ${log_xx}
> > > > lttng destroy psw-${xx} >> ${log_xx} 2>&1
> > > > echo "lttng done" >> ${log_xx}
> > > > fi
> > > >
> > > > # start yy after stopping xx
> > > > if [ -f /home/pswuser/trace/run_trace ] ; then
> > > > echo "lttng create ${yy} after closing ${xx}" > ${log_yy}
> > > > lttng create psw-${yy} \
> > > > -o /home/pswuser/trace/psw-${yy} >> ${log_yy} 2>&1
> > > > sleep 1
> > > > echo "lttng enable_event" >> ${log_yy}
> > > > lttng enable-event -a -k -s psw-${yy} >> ${log_yy} 2>&1
> > > > sleep 1
> > > > echo "lttng start" >> ${log_yy}
> > > > lttng start psw-${yy} >> ${log_yy} 2>&1
> > > > fi
> > > > # check zz
> > > > # delete unless detected or detected2 is present
> > > > if [ ! -f /home/pswuser/bin/delay-detected ] &&
> > > > [ ! -f /home/pswuser/bin/delay-detected2 ] ; then
> > > > echo "lttng delete files" >> ${log_zz}
> > > > rm -rf /home/pswuser/trace/psw-${zz}* >> ${log_zz} 2>&1
> > > > #if [ $? -eq 0 ] ; then rm ${log_zz} ; fi
> > > > fi
> > > > # delete second snapshot
> > > > if [ -f /home/pswuser/bin/delay-detected2 ] ; then
> > > > rm -f /home/pswuser/bin/delay-detected2
> > > >
> > > > # no more tracing once event detected
> > > > rm -f /home/pswuser/trace/run_trace
> > > > fi
> > > > # trigger second snapshot delete first snapshot
> > > > if [ -f /home/pswuser/bin/delay-detected ] ; then
> > > > rm -f /home/pswuser/bin/delay-detected
> > > > touch /home/pswuser/bin/delay-detected2
> > > > fi
> > > >
> > > > # ls -l /lttng-traces/psw-*
> > > > # zz (delete candidate) is old xx
> > > > echo "xx ${xx} yy ${yy} zz ${zz}"
> > > > zz=${xx}
> > > > log_zz=$log_xx
> > > >
> > > > # xx is now the new one
> > > > xx=${yy}
> > > > log_xx=$log_yy
> > > >
> > > > # check disk use
> > > > if [ -d /home/pswuser/trace ] ; then
> > > > du -h /home/pswuser/trace
> > > > fi
> > > >
> > > >
> > > > done
> > > > ----------------------------------------
> > > >
> > > > Any clues are welcome.
> > > >
> > > > Another question.
> > > >
> > > > Would it be possible to do this by setting up a ring buffer that will wrap
> > > > around after a certain time. I can easily timestamp the event I am looking
> > > > for and then capture the buffer containing the event.
> > >
> > > Stay tuned for 2013 lttng releases. This will likely be implemented this
> > > year (flight recorder and snapshots).
> > >
> > > Thanks,
> > >
> > > Mathieu
> > >
> > > >
> > > >
> > > > Regards
> > > > Phil Wilshire
> > >
> > > > _______________________________________________
> > > > lttng-dev mailing list
> > > > lttng-dev at lists.lttng.org
> > > > http://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev
> > >
> > >
> > > --
> > > Mathieu Desnoyers
> > > Operating System Efficiency R&D Consultant
> > > EfficiOS Inc.
> > > http://www.efficios.com
> > >
> > > _______________________________________________
> > > lttng-dev mailing list
> > > lttng-dev at lists.lttng.org
> > > http://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev
> >
> > --
> > Mathieu Desnoyers
> > Operating System Efficiency R&D Consultant
> > EfficiOS Inc.
> > http://www.efficios.com
> >
> > _______________________________________________
> > lttng-dev mailing list
> > lttng-dev at lists.lttng.org
> > http://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev
>
> --
> Mathieu Desnoyers
> Operating System Efficiency R&D Consultant
> EfficiOS Inc.
> http://www.efficios.com
>
> _______________________________________________
> lttng-dev mailing list
> lttng-dev at lists.lttng.org
> http://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev
--
Mathieu Desnoyers
Operating System Efficiency R&D Consultant
EfficiOS Inc.
http://www.efficios.com
More information about the lttng-dev
mailing list