linux-ext4 - Re: Corrupt ext4 fs after creation

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Tue, 14 Mar 2017 10:54:33 -0700
From:   "Darrick J. Wong" <darrick.wong@...cle.com>
To:     Daniel Schultz <d.schultz@...tec.de>
Cc:     linux-ext4@...r.kernel.org, Teresa Remmet <t.remmet@...tec.de>
Subject: Re: Corrupt ext4 fs after creation

On Tue, Mar 14, 2017 at 11:34:32AM +0100, Daniel Schultz wrote:
> Hi,
> 
> we use the Yocto Project to create custom BSPs for a AM335x SoC. After they
> upgraded the e2fsprogs from 1.42.9 to 1.43 we run into problems with our
> ext4 root file system. During the first boot systemd checks the rootfs with
> fsck.ext4 and needs a restart beacause of an error.
> 
> The fs is created with mkfs.ext4 and direct I/O and is available on ftp://ftp.phytec.de/pub/Test/phytec-headless-image-phyboard-wega-am335x-2.ext4
> 
> $ du -ks /home/schultz/yocto/PD17.1.0/build/tmp/work/phyboard_wega_am335x_2-phytec-linux-gnueabi/phytec-headless-image/1.0-r0/rootfs
>  109708
> $ truncate /var/tmp/wic/build/rootfs_root.2.ext4 -s 159674777
> $ mkfs.ext4 -F -i 8192 /var/tmp/wic/build/rootfs_root.2.ext4 -L root -d /home/schultz/yocto/PD17.1.0/build/tmp/work/phyboard_wega_am335x_2-phytec-linux-gnueabi/phytec-headless-image/1.0-r0/rootfs
>  mke2fs 1.43 (17-May-2016)
>  Discarding device blocks: done
>  Creating filesystem with 155932 1k blocks and 19520 inodes
>  Filesystem UUID: 6728344f-aa6d-4bbb-a06d-e649e36024d3
>  Superblock backups stored on blocks:
>  	8193, 24577, 40961, 57345, 73729
> 
>  Allocating group tables: done
>  Writing inode tables: done
>  Creating journal (4096 blocks): done
>  Copying files into the device: done
>  Writing superblocks and filesystem accounting information: done
> $ du -Lbks /var/tmp/wic/build/rootfs_root.2.ext4
>  155933
> 
> 
> I figured out that when I run fsck.ext4 it will perform a directory
> optimizing which leads to a non-zero error code of 1 (File system errors
> corrected). Also, I figured out that this optimizing only occurres on the
> first boot and not after creating a lot of new files and dirs. After
> checking the fs it contains more blocks than before.
> 
> root@...board-wega-am335x-2:~# fsck.ext4 -V
> e2fsck 1.43 (17-May-2016)
>         Using EXT2FS Library version 1.43, 17-May-2016
> root@...board-wega-am335x-2:~# fsck.ext4
> /dev/disk/by-id/mmc-NCard_0x2519026e-part2
> e2fsck 1.43 (17-May-2016)
> Superblock last write time (Mon Mar 13 10:53:39 2017,
>         now = Wed Jan 25 11:07:23 2017) is in the future.
> Fix<y>? yes
> Pass 1: Checking inodes, blocks, and sizes
> Pass 2: Checking directory structure
> Pass 3: Checking directory connectivity
> Pass 3A: Optimizing directories
> Pass 4: Checking reference counts
> Pass 5: Checking group summary information
> 
> root: ***** FILE SYSTEM WAS MODIFIED *****
> root: 5221/19520 files (0.9% non-contiguous), 105503/155932 blocks
> root@...board-wega-am335x-2:~# fsck.ext4
> /dev/disk/by-id/mmc-NCard_0x2519026e-part2
> e2fsck 1.43 (17-May-2016)
> root: clean, 5221/19520 files, 105503/155932 blocks
> 
> 
> Can anyone give me more informations about what's wrong with our fs and if I
> can forbid (without problems) the directory optimization?

Somewhere in that newly created image is a directory large enough to
warrant an htree index to speed up directory access.  Pass 3A is e2fsck
creating the directory index, hence the "FS was modified" message.

The htree (hash tree) indexes directory entries by hash to speed up
random directory accesses.  e2fsck can regenerate the indices, but the
rest of e2fsprogs cannot create or maintain them.  You can turn them
off, at some cost to performance.

So there are a number of options here -- (1) reduce directory sizes,
which might not be feasible.

(2) You could increase the block size since block size == page size has
less runtime overhead... unless you really do need to have a 155MB
image with as little slack as possible.

(3) You could turn off directory indexing (mkfs -O ^dir_index) which
removes the fsck surprise but also makes directory access much slower.

(4) Run e2fsck immediately after mkfs so that the directory
optimizations are baked into the root image.

I'd probably do #4 and/or #2, personally.

--D

> 
> Thanks
> 
> -- 
> Mit freundlichen Grüßen,
> With best regards,
>   Daniel Schultz