Strange UFS write problem & SU+J "unexpected inconsistences" on 9.1-STABLE r253105 after it on OTHER filesystems.

[Kirk McKusick]

> Date: Wed, 25 Sep 2013 23:32:48 +0400
> From: Lev Serebryakov <lev at FreeBSD.org>
> To: Kirk McKusick <mckusick at mckusick.com>
> CC: Jeff Roberson <jroberson at jroberson.net>,
>         freebsd-fs <freebsd-fs at FreeBSD.org>
> Subject: Re: Strange UFS write problem & SU+J "unexpected inconsistences" on 9.1-STABLE r253105 after it on OTHER filesystems.
> 
> Hello, Kirk.
> You wrote 23 =D1=81=D0=B5=D0=BD=D1=82=D1=8F=D0=B1=D1=80=D1=8F 2013 =D0=B3.,=
>  2:24:02:
> 
> 
> KM> Have you run a manual (fsck -f) on the affected filesystems? If so,
> KM> was it able to clean them up? If not, please do so and save the output
> KM> of fsck (using script command is usually the easiest way to do this).
>  I've stopped server and checked two FSes with WTITE errors (/ and /usr)
> with fsck -f, but it didn't find any errors at all.
>  I have / dumped (completely, as binary blob) and could upload it when I
> find and mask all passwords :)
> 
> --=20
> // Black Lion AKA Lev Serebryakov <lev at FreeBSD.org>

I do not think that we are likely to find out anything useful from
looking at your dumpped /. Notably the errors that were produced were
because an indirect block had gottened trashed. Looking at the file
image is not going to tell us how they got trashed. Usual cases are
memory errors, address line errors on the I/O bus, or I/O error in the
disk itself. The journal does not know of these errors, so does not
look to correct them. Hence the need for a manual fsck.

	Kirk McKusick