[lvm-devel] lvresize in test mode sends BLKDISCARD to SSD block devices, this will result in the file system corruption

[Zdenek Kabelac]

Dne 22.6.2018 v 11:31 Gang He napsal(a):
> Hello List,
> 
> I found a bug for lvresize command, this issue can be reproduced by lvm2-2.02.178 code base.
> The reproduce steps are as below,


Hi

Good catch - yep - it seem in this case  ioctl operation are flowing to device 
even in test mode.

Please create BZ for lvm2 on this issue.


Zdenek


> 1) create a LV on a SSD based partition, then format this LV with ext4 file system, and mount this LV.
> e.g.
> linux-cy2k:/ # lsblk
> ...
> sdc                               8:32   0  37.3G  0 disk
> ├─sdc1                            8:33   0     2G  0 part [SWAP]
> ├─sdc2                            8:34   0  14.7G  0 part
> │ └─cluster--vg1-test--lv       254:20   0    14G  0 lvm  /mnt/ext4  <<= this LV
> └─sdc3                            8:35   0  20.6G  0 part
> 
> 2) enable issue discards to a logical volumes's underlying physical volume via edit "/etc/lvm/lvm.conf" file.
> e.g.
> linux-cy2k:/ # vi /etc/lvm/lvm.conf
> set "issue_discards = 1"
> 
> 3) run lvresize command in test mode.
> e.g.
> inux-cy2k:/ # lvresize -v -n -L -10GB /dev/cluster-vg1/test-lv --test
> 
> 4) unmount this LV, then mount this LV again.
> You will probably encounter the file system on this LV is corrupted.
> e.g.
> linux-cy2k:~ # mount /dev/cluster-vg1/test-lv /mnt/ext4/
> mount: wrong fs type, bad option, bad superblock on /dev/mapper/cluster--vg1-test--lv,
>         missing codepage or helper program, or other error
> 
>         In some cases useful info is found in syslog - try
>         dmesg | tail or so.
> 
> linux-cy2k:~ # dmesg
> [3711112.342424] JBD2: no valid journal superblock found
> [3711112.342429] EXT4-fs (dm-20): error loading journal
> 
> To summarize, I think this behavior is a bug and is very dangerous,
> the command in test mode should not send BLKDISCARD to the underlying block device,
> the BLKDISCARD command will make the related sectors on the block device be trimmed (lost data).
> 
> 
> Thanks
> Gang
>   
> 
> --
> lvm-devel mailing list
> lvm-devel at redhat.com
> https://www.redhat.com/mailman/listinfo/lvm-devel
>