Re: panic: ffs_blkfree: freeing free block

On Wednesday 07 November 2007, Ken Smith wrote:
> On Tue, 2007-10-23 at 11:35 +0200, Arjan van Leeuwen wrote:
> > 2007/10/19, Tim Bishop <tim-lists_at_bishnet.net>:
> > > On Thu, Oct 18, 2007 at 06:40:04PM +0200, Arjan van Leeuwen wrote:
> > > > > are softupdates on ?
> > > > >
> > > > > Yes. Turning them off seems to fix the problem (fingers crossed - I
> > >
> > > only
> > >
> > > > > turned it off this morning, didn't have a panic yet)
> > > >
> > > > After working for a whole day without softupdates, I can say that
> > >
> > > turning
> > >
> > > > them off at least causes less panics to happen than with them turned
> > > > on, maybe they even don't happen at all without softupdates; I
> > > > haven't had a panic all day, while I have one every few hours with
> > > > softupdates turned
> > >
> > > on.
> > >
> > > I've been running for the best part of a day now with softupdates
> > > turned off and so far no panics. I'm running tinderbox on the host, and
> > > it would quite reliably crash it before.
> > >
> > > Of course it's hard to say if this has fixed the problem... maybe it
> > > doesn't happen as often, or maybe my data is being slowly chewed up
> > > instead ;-)
> >
> > It looks like on the same system, I'm able to reliably panic zfs as well,
> > under the exact same conditions (i.e. linking a particularly big piece of
> > software). Maybe this is not a problem in the filesystem at all. I've not
> > been able to get a coredump yet from the zfs panic.
> > This is in 7.0-PRERELEASE, btw (I switched to RELENG_7 when the branching
> > happened).
>
> Is anyone who had been able to trigger this panic still having problems
> with recent kernels (and soft updates turned on)?  I've checked with a
> few people who had been experiencing the panic and they can no longer
> trigger it.
>
> It's at least a tiny bit possible some of the VM fixes that have gone in
> addressed this problem.  We'd like to find out if anyone can still
> trigger this.
I can no longer reproduce it any more.

-
Pieter