Re: Panic in ZFS: solaris assert: dn->dn_datablkshift != 0

On Sun, Sep 08, 2013 at 11:14:44PM +0200, Jeremie Le Hen wrote:
> On Sun, Sep 08, 2013 at 10:05:55PM +0100, Steven Hartland wrote:
> > > On Sun, Sep 08, 2013 at 09:02:48PM +0100, Steven Hartland wrote:
> > >> > On Sun, Sep 08, 2013 at 03:17:09PM +0100, Steven Hartland wrote:
> > >> >> I believe this was added by this change set:-
> > >> >> http://svnweb.freebsd.org/base?view=revision&revision=253821
> > >> >> 
> > >> >> Might want to try back out that change and see if everything
> > >> >> works after that?
> > >> > 
> > >> > Actually, I already rolled back my kernel to August 1st:
> > >> > 
> > >> > # svn info .
> > >> > Path: .
> > >> > Working Copy Root Path: /usr/src
> > >> > URL: http://svn0.us-west.freebsd.org/base/head/sys
> > >> > Repository Root: http://svn0.us-west.freebsd.org/base
> > >> > Repository UUID: ccf9f872-aa2e-dd11-9fc8-001c23d0bc1f
> > >> > Revision: 253847
> > >> > Node Kind: directory
> > >> > Schedule: normal
> > >> > Last Changed Author: ian
> > >> > Last Changed Rev: 253847
> > >> > Last Changed Date: 2013-07-31 21:14:00 +0200 (Wed, 31 Jul 2013)
> > >> > 
> > >> > 
> > >> > And the problem seems to have gone away.  I could perform a full zfs
> > >> > send/receive whereas it would trigger a panic 100% of the time with a
> > >> > recent kernel.
> > >> 
> > >> I still think r253821 is the cause the reason being is prior to r253996
> > >> ASSERTS in ZFS where not actually active in HEAD.
> > >> 
> > >> So if you could roll forward but then backout r253821 and confirm this
> > >> is indeed the cause that would be a good starting point.
> > >> 
> > >> If this is indeed the cause be worth engaging Matthew Ahrens cc'ed to
> > >> find out the reasoning behind the new ASSERT why you may be hitting it?
> > > 
> > > 
> > 
> > Errm, was there meant to be some content in your reply Jeremie, as it seems
> > to be missing?
> 
> Indeed, probably a bad key combo in vi :).
> 
> I'm reverting r253821 and r254753 (the second one was supposingly fixing
> the first one) and recompiling my kernel.
> 
> I will let you know.

So far so good, I've been able to synchronize my datasets beyond the
point where it crashed last time.

Matthew, do you have any idea of a fix I could try on top of FreeBSD's
r253821 and r254753?

-- 
Jeremie Le Hen

Scientists say the world is made up of Protons, Neutrons and Electrons.
They forgot to mention Morons.