Date: Mon, 27 Oct 2014 22:50:14 -0400 From: Scott Long <scottl@samsco.org> To: Dominic Fandrey <kamikaze@bsdforen.de> Cc: Alexander Motin <mav@freebsd.org>, freebsd-stable@freebsd.org Subject: Re: File system issues Message-ID: <ABBC227A-F1FD-4FF5-93B9-4AAB2613CE71@samsco.org> In-Reply-To: <544BC863.2040607@bsdforen.de> References: <544BC863.2040607@bsdforen.de>
next in thread | previous in thread | raw e-mail | index | archive | help
> On Oct 25, 2014, at 11:57 AM, Dominic Fandrey <kamikaze@bsdforen.de> = wrote: >=20 > Two or 3 days ago, after an update to stable/10 my UFS file system > started acting weird. I have freezes and files disappearing from the > system. >=20 > The first time that happened SU+J failed me. I.e. I went into single > user mode and on the second fsck run it would still find errors that > weren't successfully corrected due to the journal. That was the first > time with the Samsung 840 PRO SSD. >=20 > I rebuilt kernel/world, turned off journaling and rebooted, but the > problems persisted. Yesterday I updated again: > FreeBSD AprilRyan.norad 10.1-PRERELEASE FreeBSD 10.1-PRERELEASE #0 = r273588: Fri Oct 24 17:18:14 CEST 2014 = root@AprilRyan.norad:/usr/obj/S403/amd64/usr/src/sys/S403 amd64 >=20 > Today I suddenly couldn't use sysctl any more. It turned out /sbin was > suddenly empty. A couple of minutes later the system froze. After the > hard reset it came up like this (shortened to relevant bits): > http://pastebin.com/kXLtg3JR >=20 > As you can see I'm using geli for everything but /boot. After a hard > reset I usually go into single user mode and run fsck twice (ever = since > I had some really bad experiences). >=20 > It doesn't come up clean, but what turns up in lost+found are just > files from the browser cache and /sbin is back. >=20 > What terrifies me (apart from disappearing files) is what happens > after the system goes through a suspend and resume cycle, starting > at line 258 in the pastebin. The harddisk comes back up with new > errors! >=20 > I cannot pin down which version was the one last working for me, but > it's uname already said something about 10.1-PRERELEASE. >=20 Your symptoms sound very much like what we experienced with a recent merge of code to 10-STABLE. We don=E2=80=99t completely understand the = problem, but backing out r273272 avoids the problem, though it=E2=80=99s = suspected that there=E2=80=99s a bigger problem. r273638 in HEAD backs out everything suspect, and = we=E2=80=99ve been running for several days now with no problems with it. I believe = that it=E2=80=99ll be committed to 10-STABLE in the next day or two. Scott
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?ABBC227A-F1FD-4FF5-93B9-4AAB2613CE71>