Thread (13 messages) 13 messages, 2 authors, 2016-09-01

Re: btrfs-progs 4.7, check reports many "incorrect local backref count" messages

From: Chris Murphy <hidden>
Date: 2016-09-01 23:59:33

On Thu, Sep 1, 2016 at 5:51 PM, Paul Jones [off-list ref] wrote:
quoted
-----Original Message-----
From: chris@colorremedies.com [mailto:chris@colorremedies.com] On

On Thu, Sep 1, 2016 at 12:51 AM, Paul Jones [off-list ref] wrote:
quoted
quoted
-----Original Message-----
From: linux-btrfs-owner@vger.kernel.org [mailto:linux-btrfs-
owner@vger.kernel.org] On Behalf Of Chris Murphy
Sent: Thursday, 1 September 2016 7:59 AM
To: Btrfs BTRFS <redacted>
Subject: Re: btrfs-progs 4.7, check reports many "incorrect local
backref count" messages

This is still happening with btrfs-progs 4.7.1 and there is zero
information in the long result what to do about the problem, and
whether it's sane to try have --repair fix it, let alone what the original
cause of the problem was.
quoted
I just potentially damaged a perfectly good filesystem because of this. I was
getting hundreds of "Incorrect local backref count" so I decided to try repair,
which seemed to complete ok. I then rescanned without repair and btrfs
check eventually crashed with an assertion. That's when I figured something
may be wrong.

Wait, so you did a --repair and then the following-up check crashes?
But does the file system mount and does it still work? The first is bad enough,
but if it won't mount this is terrible.
Correct. I forgot to mention the crash was with btrfs-progs 4.7.1 and kernel 4.7.2. After that I reverted to btrfs-progs 4.6.1 and kernel 4.6.7
The filesystem still mounted, so I tried to run scrub but that got stuck at 0 bytes (with no errors). I reset the system (reboot didn't finish) and run repair a few more times, and even though there were still errors every time I noticed the last few errors appeared to be the same.
I mounted and tried scrub again, which worked the second time. I let it run overnight and it didn't find any errors. I'm just about to rsync (with checksums) from the backup to verify the data is still good.

I'd suggest btrfs-progs 4.7.x is withdrawn until this can be fixed.
Yeah upstream can't really do that, it's up to the distros. The main
thing though is that it didn't actually make the file system worse
off, it just sounds like it wasted some time on your end.

My complaint about this is I have no idea what the messages mean, how
big of a problem they are, whether they should be repaired, and what
to do about the fact they claim to be repaired but aren't.



-- 
Chris Murphy
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help