Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 12 Mar 2015 05:18:17 +0000
From:      Glen Barber <gjb@FreeBSD.org>
To:        freebsd-stable@FreeBSD.org
Cc:        FreeBSD Release Engineering Team <re@FreeBSD.org>
Subject:   HEADS-UP: system deadlock issues when rebooting after updating or upgrading to 10.1-RELEASE
Message-ID:  <20150312051817.GI51985@hub.FreeBSD.org>

next in thread | raw e-mail | index | archive | help

--+ZmrHH5cGjskQnY1
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline

It has become apparent that issues reported in PR 195458 [1] do not
apply to freebsd-update(8) users alone, and that the deadlock on reboot
(either via reboot(8) or shutdown(8) with the '-r' flag) can be reliably
reproduced under certain conditions.

Specifically, the issue can be reliably reproduced on systems using UFS
as the root filesystem with soft-updates and journaling enabled (a.k.a.,
SU+J).

Systems using ZFS or UFS without SU+J appear to be unaffected at this
point (noted by various people in the referenced PR, as well as my own
tests).  Based on the comments in the referenced PR, it is unclear if
this deadlock can occur under other conditions, but it seems to be
contained to UFS with SU+J.  If you have experienced a deadlock after
upgrading a 10.1-RELEASE system, please update the referenced PR with
pertinent information (filesystem, and filesystem options, etc.).  While
the underlying cause, and more importantly, a resolution is being
investigated, it is critical we have as much information as possible.

Once the underlying issue is identified, and we have a tested fix for
the problem, an EN will be issued as soon as reasonably possible.

To be clear, this issue manifests as an unresponsive system after
issuing a reboot(8) (or in some cases, a shutdown(8)) command, after the
"All buffers synced." message is displayed, and this heads-up is
addressing only those cases (or potential cases).  All other filesystem
related should be reported in a separate PR, as usual.

Having said all of the above, I as the RE for 10.1-RELEASE, personally
apologize to everyone that has run into this issue.  I am still baffled
why I did not run into this problem when testing 10.1-RELEASE (or any of
the -BETA or -RC builds).  I can only attribute this to a change in my
testing environment (i.e., updated hypervisor software), but given the
details and reproducibility after recent testing, am unable to explain
how this was not discovered earlier.

[1] https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=195458

Glen
On behalf of:	re@


--+ZmrHH5cGjskQnY1
Content-Type: application/pgp-signature

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2
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=Klgm
-----END PGP SIGNATURE-----

--+ZmrHH5cGjskQnY1--



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20150312051817.GI51985>