ZFS Crash

Johan Hendriks Johan at double-l.nl
Tue May 26 14:24:44 UTC 2009

>On 26.05.09 14:04, Larry Rosenman wrote:
>> On Mon, 25 May 2009, Larry Rosenman wrote:
>>> On Mon, 25 May 2009, Larry Rosenman wrote:
>>>> after looking at the code, never mind the "don't call doadump", so we'll
>>>> get the textdump.
>>>> Thanks rwatson for the textdump stuff!
>>> Here is current stats before we crash. Does any of this look totally
>>> out of line?
>> It crashed again, but did *NOT* make it into ddb enough to do the textdump.
>> It was hung with the backtrace (looks like the same, but I couldn't
>> scroll the screen back).
>> Ideas?
>> I'm really concerned that there is a problem.

>Hi i guess I'm having a similar/the same problem. The panic looks the 
>same, and the trace is a little different, I've had a lot of these 
>panics on this machine, and the stack trace looked different quite often.

>The panic and show uma and show malloc can be found here: 

>I have a coredump written to swap, but this panic is very easy to 
>trigger and it panics every time savecore runs... It is also easy to 
>trigger this with a buildworld -j5.

>This is a quadcore Xeon with a a slow SATA disk and only 1GB of mem. 
>(It's only a test machine, so i can test any suggestion.) This box used 
>to run stable with kmem max set to 768M.

>Anything i can do?


I have several panics also i am using cvsup-mirror on a zfs raidz2.
It did run good for quite some time, but last week after a buildworld it started to panic as soon as /usr/local/etc/cvsup/update.sh was running.
I had vm.kmem_size="512M" and vm.kmem_size_max="512M".
I had a custom kernel with some options added and some taken off.
The box is a HP Proliant 110 with 1GB memory and is running the amd64 version of FreeBSD.  

My custom kernel had the following.

# Debugging for use in -current
nooptions   KDB         # Enable kernel debugger support.
nooptions   DDB         # Support DDB.
nooptions   GDB         # Support remote GDB.
nooptions   INVARIANTS      # Enable calls of extra sanity checking
nooptions   INVARIANT_SUPPORT   # Extra sanity checks of internal structures, required by INVARIANTS
nooptions   WITNESS         # Enable checks to detect deadlocks and cycles
nooptions   WITNESS_SKIPSPIN    # Don't run witness on spinlocks for speed

options     DEVICE_POLLING

# pf
options     ALTQ
options     ALTQ_CBQ
options     ALTQ_RED
options     ALTQ_RIO
options     ALTQ_HFSC
options     ALTQ_CDNR
options     ALTQ_PRIQ
device      pf
device      pflog
device      pfsync

I did a new buildworld with a GENERIC kernel with pf part still there. But now i can not panic it again.
So maybe it is the options     DEVICE_POLLING option.

i will try to add that again to the kernel and try to panic it again.
Do you have the DEVICE_POLLING option in your kernel?

BTW i am no FreeBSD or debug expert.
This is just an observation.

Johan Hendriks

No virus found in this outgoing message.
Checked by AVG - www.avg.com 
Version: 8.5.339 / Virus Database: 270.12.39/2134 - Release Date: 05/25/09 18:14:00

More information about the freebsd-current mailing list