Bug #997

vkernel livelocked and core dumped when killed on -HEAD

Added by rumcic about 6 years ago. Updated almost 5 years ago.

Status:ClosedStart date:
Priority:NormalDue date:
Assignee:-% Done:

0%

Category:-
Target version:-

Description

Tried to cvs checkout the pkgsrc tree and managed to get "intr 1 at 40001/40000
hz, livelocked limit engaged!". After kill-ing it, it core dumped. The
backtrace repeats itself over and over, I uploaded kernel and kernel.core to
leaf:~rumko/crash/vkernel/30.04.2008/

Part of the backtrace:
#0 exc_segfault (signo=11, info=0xbf7fd2f8, ctxp=0xbf7fd018)
at /usr/src/sys/platform/vkernel/i386/exception.c:140
#1 <signal handler called>
#2 kern_trap (frame=0xbf7fd38c)
at /usr/src/sys/platform/vkernel/i386/trap.c:656
#3 0x081be473 in exc_segfault (signo=11, info=0xbf7fd658, ctxp=0xbf7fd378)
at /usr/src/sys/platform/vkernel/i386/exception.c:150
#4 <signal handler called>
#5 kern_trap (frame=0xbf7fd6ec)
at /usr/src/sys/platform/vkernel/i386/trap.c:656
#6 0x081be473 in exc_segfault (signo=11, info=0xbf7fd9b8, ctxp=0xbf7fd6d8)
at /usr/src/sys/platform/vkernel/i386/exception.c:150
#7 <signal handler called>
#8 kern_trap (frame=0xbf7fda4c)
at /usr/src/sys/platform/vkernel/i386/trap.c:656
#9 0x081be473 in exc_segfault (signo=11, info=0xbf7fdd18, ctxp=0xbf7fda38)
at /usr/src/sys/platform/vkernel/i386/exception.c:150
#10 <signal handler called>
#11 kern_trap (frame=0xbf7fddac)
at /usr/src/sys/platform/vkernel/i386/trap.c:656
#12 0x081be473 in exc_segfault (signo=11, info=0xbf7fe078, ctxp=0xbf7fdd98)
at /usr/src/sys/platform/vkernel/i386/exception.c:150
#13 <signal handler called>
#14 kern_trap (frame=0xbf7fe10c)
at /usr/src/sys/platform/vkernel/i386/trap.c:656
#15 0x081be473 in exc_segfault (signo=11, info=0xbf7fe3d8, ctxp=0xbf7fe0f8)
at /usr/src/sys/platform/vkernel/i386/exception.c:150
#16 <signal handler called>
#17 kern_trap (frame=0xbf7fe46c)
at /usr/src/sys/platform/vkernel/i386/trap.c:656
#18 0x081be473 in exc_segfault (signo=11, info=0xbf7fe738, ctxp=0xbf7fe458)
at /usr/src/sys/platform/vkernel/i386/exception.c:150
#19 <signal handler called>
#20 kern_trap (frame=0xbf7fe7cc)
at /usr/src/sys/platform/vkernel/i386/trap.c:656
#21 0x081be473 in exc_segfault (signo=11, info=0xbf7fea98, ctxp=0xbf7fe7b8)
at /usr/src/sys/platform/vkernel/i386/exception.c:150
#22 <signal handler called>
#23 kern_trap (frame=0xbf7feb2c)
at /usr/src/sys/platform/vkernel/i386/trap.c:656
#24 0x081be473 in exc_segfault (signo=11, info=0xbf7fedf8, ctxp=0xbf7feb18)
at /usr/src/sys/platform/vkernel/i386/exception.c:150
#25 <signal handler called>
#26 kern_trap (frame=0xbf7fee8c)
at /usr/src/sys/platform/vkernel/i386/trap.c:656
#27 0x081be473 in exc_segfault (signo=11, info=0xbf7ff158, ctxp=0xbf7fee78)
at /usr/src/sys/platform/vkernel/i386/exception.c:150
#28 <signal handler called>
#29 kern_trap (frame=0xbf7ff1ec)
at /usr/src/sys/platform/vkernel/i386/trap.c:656
#30 0x081be473 in exc_segfault (signo=11, info=0xbf7ff4b8, ctxp=0xbf7ff1d8)
at /usr/src/sys/platform/vkernel/i386/exception.c:150
#31 <signal handler called>
#32 kern_trap (frame=0xbf7ff54c)
at /usr/src/sys/platform/vkernel/i386/trap.c:656
#33 0x081be473 in exc_segfault (signo=11, info=0xbf7ff818, ctxp=0xbf7ff538)
at /usr/src/sys/platform/vkernel/i386/exception.c:150
#34 <signal handler called>
#35 kern_trap (frame=0xbf7ff8ac)
at /usr/src/sys/platform/vkernel/i386/trap.c:656
#36 0x081be473 in exc_segfault (signo=11, info=0xbf7ffb78, ctxp=0xbf7ff898)
at /usr/src/sys/platform/vkernel/i386/exception.c:150
#37 <signal handler called>
#38 kern_trap (frame=0xbf7ffc0c)
at /usr/src/sys/platform/vkernel/i386/trap.c:656
#39 0x081be473 in exc_segfault (signo=11, info=0xbf7ffed8, ctxp=0xbf7ffbf8)
at /usr/src/sys/platform/vkernel/i386/exception.c:150
#40 <signal handler called>
#41 kern_trap (frame=0xbf7fff6c)
at /usr/src/sys/platform/vkernel/i386/trap.c:656
#42 0x081be473 in exc_segfault (signo=11, info=0xbf800238, ctxp=0xbf7fff58)
at /usr/src/sys/platform/vkernel/i386/exception.c:150
#43 <signal handler called>
#44 kern_trap (frame=0xbf8002cc)
at /usr/src/sys/platform/vkernel/i386/trap.c:656
#45 0x081be473 in exc_segfault (signo=11, info=0xbf800598, ctxp=0xbf8002b8)
--
Regards,
Rumko

History

#1 Updated by dillon about 6 years ago

:Tried to cvs checkout the pkgsrc tree and managed to get "intr 1 at 40001/40000
:hz, livelocked limit engaged!". After kill-ing it, it core dumped. The
:backtrace repeats itself over and over, I uploaded kernel and kernel.core to
:leaf:~rumko/crash/vkernel/30.04.2008/

Did the vkernel actually lock-up or was it still responding? e.g. did
you try ^Z'ing or ^C'ing the checkout? It is quite possible that
the IO co-thread actually managed to exceed the livelock limit without
it being a bug.

The crash was due to the IO co-thread crashing. The IO co-thread took
the shutdown signal from the vkernel when you killed it, but since it
doesn't have a globaldata context that blew it up.

I will mask the shutdown signal on the co-thread to fix the core
dumping issue but we still need to investigate the livelock. Have
you been able to reproduce it?

-Matt

#2 Updated by rumcic about 6 years ago

Matthew Dillon wrote:

If I remember correctly, it was completely locked-up, but I will retry tonight.
--
Regards,
Rumko

#3 Updated by rumcic about 6 years ago

Rumko wrote:

Managed to lock up the host machine by dd-ing a large (10GB) file ... but
couldn't get a memory dump (it seems it did not create it) ... I'll continue
with this tomorrow when I'll feel bolder
--
Regards,
Rumko

#4 Updated by dillon about 6 years ago

:Managed to lock up the host machine by dd-ing a large (10GB) file ... but
:couldn't get a memory dump (it seems it did not create it) ... I'll continue
:with this tomorrow when I'll feel bolder
:--
:Regards,
:Rumko

I can temporarily lock up a machine doing that, but it recovers. I
haven't tracked it down yet.

Try to determine whether your dd is permanently locking up the box, or
just locking it up until the dd ends.

-Matt
Matthew Dillon
<>

#5 Updated by tuxillo almost 5 years ago

Is this still reproducible?

#6 Updated by tuxillo almost 5 years ago

Sorry I forgot to say that I've checked out pkgsrc inside a vkernel many times
without problems.

#7 Updated by dillon almost 5 years ago

The older vkernel related issues can probably be closed. There have
been massive changes since 2008.

-Matt
Matthew Dillon
<>

Also available in: Atom PDF