Re: A file cloned with "cp --reflink" different from the original one?

[Jian Lin]

Thanks. I will try the new version soon.

2009/11/23 Pádraig Brady <address@hidden>:
> Jian Lin wrote:
>> 2009/11/23 Pádraig Brady <address@hidden>:
>>> Jian Lin wrote:
>>>> I installed BtrFS 0.19 and GNU coreutils 8.1 on my Ubuntu 9.10.
>>>> I tried to clone some files with "cp --reflink" to make them 
>>>> "copy-on-write".
>>>> However, I found some of the files cloned have different MD5s to the
>>>> original one.
>>>>
>>>> Is BtrFS (or cp with reflink) buggy?
>>>> Or it is indeed a feature that I used incorrectly?
>>>>
>>>> Thanks!
>>>>
>>>> address@hidden:/mnt# uname -a
>>>> Linux lj-laptop 2.6.31-15-generic #50-Ubuntu SMP Tue Nov 10 14:54:29
>>>> UTC 2009 i686 GNU/Linux
>>>>
>>>> address@hidden:/mnt# mount | grep btrfs
>>>> /dev/sda4 on /mnt type btrfs (rw)
>>>>
>>>> address@hidden:/mnt# cp --version
>>>> cp (GNU coreutils) 8.1
>>>>
>>>> address@hidden:/mnt# cd WinXP_CHS/
>>>> address@hidden:/mnt/WinXP_CHS# cp WinXP_CHS.vmdk WinXP_CHS-ref.vmdk 
>>>> --reflink
>>>>
>>>> address@hidden:/mnt/WinXP_CHS# ls -la WinXP_CHS.vmdk WinXP_CHS-ref.vmdk
>>>> -rwxr-xr-x 1 root root 1713766400 Nov 23 15:11 WinXP_CHS-ref.vmdk
>>>> -rwxr-xr-x 1 root root 1713766400 Nov 23 13:32 WinXP_CHS.vmdk
>>>>
>>>> address@hidden:/mnt/WinXP_CHS# md5sum WinXP_CHS.vmdk WinXP_CHS-ref.vmdk
>>>> f6da592d056167a44b089d70fa46f863  WinXP_CHS.vmdk
>>>> e3ee0c2c17771811c80eed088c20987d  WinXP_CHS-ref.vmdk
>>> Yikes. Are you 100% sure nothing is changing those vm files?
>>
>> I think so. I run md5sum immediately after cloning the file.
>>
>>> Note I wouldn't 100% trust the mtime, so an md5sum before
>>> and after would be good to confirm.
>>>
>>> You could try this simple prog to do the clones to
>>> implicate coreutils or otherwise:
>>>
>>> #include <sys/ioctl.h>
>>> #include <sys/types.h>
>>> #include <sys/stat.h>
>>> #include <fcntl.h>
>>> #include <stdio.h>
>>> #define BTRFS_IOC_CLONE 1074041865
>>>
>>> int main(int argc, const char **argv)
>>> {
>>>  int in = open(argv[1], O_RDONLY);
>>>  int out = open(argv[2], O_CREAT|O_TRUNC|O_WRONLY, 0644);
>>>  if (ioctl(out, BTRFS_IOC_CLONE, in)==-1) {
>>>      fprintf(stderr, "ioctl error = %m\n");
>>>  }
>>>  close(in);
>>>  close(out);
>>>  return 0;
>>> }
>>
>> I compiled and tried this program.
>> ./clone WinXP_CHS.vmdk WinXP_CHS-clone.vmdk
>> then I got md5 of WinXP_CHS-clone.vmdk, which is the same as that of
>> WinXP_CHS-ref.vmdk (made by cp --reflink), and different from the md5
>> of original WinXP_CHS.vmdk.
>>
>> The cloned vm cannot run correctly, which said "hal.dll lost". The
>> cloned Linux vm I mentioned before also reported "/bin/bash lost".
>> Another tiny linux vm I cloned with the same md5 to its original vmdk
>> can run correctly.
>
> OK then it sounds like an issue with BTRFS (CC'd)
> (which could already be fixed since it's in active development).
>
> cheers,
> Pádraig.
>



-- 

林健
中国科学院计算技术研究所网络重点实验室
北京市海淀区科学院南路6号 邮编：100190

Jian LIN
Key Laboratory of Network Science and Technology,
Institute of Computing Technology,
Chinese Academy of Sciences,
P.O.Box 2704, Beijing 100190, China.

Website:    http://www.linjian.org
Blog:       http://blog.linjian.org
E-mail/IM:  address@hidden