Re: ext4 error when testing virtio-scsi & vhost-scsi

[Zhangfei Gao <zhangfei.gao@gmail.com>]

Dear Ted

On Wed, Jul 13, 2016 at 12:43 AM, Theodore Ts'o <tytso@mit.edu> wrote:
> On Tue, Jul 12, 2016 at 03:14:38PM +0800, Zhangfei Gao wrote:
>> Some update:
>>
>> If test with ext2, no problem in iblock.
>> If test with ext4, ext4_mb_generate_buddy reported error in the
>> removing files after reboot.
>>
>>
>> root@(none)$ rm test
>> [   21.006549] EXT4-fs error (device sda): ext4_mb_generate_buddy:758: group 18
>> , block bitmap and bg descriptor inconsistent: 26464 vs 25600 free clusters
>> [   21.008249] JBD2: Spotted dirty metadata buffer (dev = sda, blocknr = 0). Th
>> ere's a risk of filesystem corruption in case of system crash.
>>
>> Any special notes of using ext4 in qemu?
>
> Ext4 has more runtime consistency checking than ext2.  So just because
> ext4 complains doesn't mean that there isn't a problem with the file
> system; it just means that ext4 is more likely to notice before you
> lose user data.
>
> So if you test with ext2, try running e2fsck afterwards, to make sure
> the file system is consistent.
>
> Given that I'm reguarly testing ext4 using kvm, and I haven't seen
> anything like this in a very long time, I suspect the problemb is with
> your SCSI code, and not with ext4.
>

Do you know what's the possible reason of this error.

Have tried 4.7-rc2, same issue exist.
It can be reproduced by fileio and iblock as backstore.
It is easier to happen in qemu like this process:
qemu-> mount-> dd xx -> umout -> mount -> rm xx, then the error may
happen, no need to reboot.

ramdisk can not cause error just because it just malloc and memcpy,
while not going to blk layer.

Also tried creating one file in /tmp, used as fileio, also can reproduce.
So no real device is based.

like:
cd /tmp
dd if=/dev/zero of=test bs=1M count=1024; sync;
targetcli
#targetcli
(targetcli) /> cd backstores/fileio
(targetcli) /> create name=file_backend file_or_dev=/tmp/test size=1G
(targetcli) /> cd /vhost
(targetcli) /> create wwn=naa.60014052cc816bf4
(targetcli) /> cd naa.60014052cc816bf4/tpgt1/luns
(targetcli) /> create /backstores/fileio/file_backend
(targetcli) /> cd /
(targetcli) /> saveconfig
(targetcli) /> exit

/work/qemu.git/aarch64-softmmu/qemu-system-aarch64 \
    -enable-kvm -nographic -kernel Image \
    -device vhost-scsi-pci,wwpn=naa.60014052cc816bf4 \
    -m 512 -M virt -cpu host \
    -append "earlyprintk console=ttyAMA0 mem=512M"

in qemu:
mkfs.ext4 /dev/sda
mount /dev/sda /mnt/
sync; date; dd if=/dev/zero of=/mnt/test bs=1M count=100; sync; date;

using dd test, then some error happen.
log like:
oot@(none)$ sync; date; dd if=/dev/zero of=test bs=1M count=100; sync;; date;
[ 1789.917963] sbc_parse_cdb cdb[0]=0x35
[ 1789.922000] fd_execute_sync_cache immed=0
Tue Jul 19 07:26:12 UTC 2016
[  200.712879] EXT4-fs error (device sda) [ 1790.191770] sbc_parse_cdb
cdb[0]=0x2a
in ext4_reserve_inode_write:5362[ 1790.198382]  fd_execute_rw
: Corrupt filesystem
[  200.729001] EXT4-fs error (device sda) [ 1790.207843] sbc_parse_cdb
cdb[0]=0x2a
in ext4_reserve_inode_write:5362[ 1790.214495]  fd_execute_rw
: Corrupt filesystem

Looks like the error usually happens after SYCHRONIZE CACHE, but not
for sure it is always happen after sync cache.

Thanks