[OpenAFS-devel] openafs-1.2.6 namei-salvager destroys volumes
Hartmut Reuter
reuter@rzg.mpg.de
Wed, 28 Aug 2002 10:47:26 +0200
Sorry, the correct name is AFS_64BIT_IOPS_ENV !
Hartmut Reuter wrote:
> In the meantime I remember: I had to set AFS_64BIT_IOPSW. Without it the
> macros VN_GET_INO and VN_SET_INO etc. do not fill the high order bits of
> Inode. Many architectures have this already set in param.<arch>.h with
> the comment that it is needed for the namei-interface.
>
> Hartmut
>
> Hartmut Reuter wrote:
>
>>
>> I remember such effects. The names you find by find are much too
>> short. Looks like the high order half of your Inode fileds were not
>> properly filled or they are not interpreted as int64.
>>
>> Hartmut Reuter
>>
>> Michael Niksch wrote:
>>
>>> I compiled openafs-1.2.6 on AIX 4.3.3 with the
>>> --enable-namei-fileserver flag. While the fileserver itself seems to be
>>> working, running salvager destroys the data. The underlying /vicepa
>>> filesystem is a standard JFS created by AIX 4.3.3.
>>>
>>> Any ideas what might be going wrong would be highly appreciated.
>>>
>>> Below is a demonstration of the problem, which occurs even on a newly
>>> created, empty volume. Interestingly enough, no visible files in
>>> /vicepa are deleted, nor does the number of used inodes change during
>>> the run of salvager. Nevertheless, the volume is no longer accessible.
>>> With actual data in the volume, I have even seen the fileserver refuse
>>> to attach the volume at all after salvaging.
>>>
>>> # find /vicepa
>>> /vicepa
>>> /vicepa/lost+found
>>> /vicepa/Lock
>>> /vicepa/Lock/vicepa
>>> /vicepa/AFSIDat
>>> /vicepa/AFSIDat/README
>>> # df /vicepa Filesystem 512-blocks Free %Used Iused
>>> %Iused Mounted on
>>> /dev/lv02 262144 253808 4% 21 1% /vicepa
>>> # vos create vorab /vicepa test Volume 536870986 created on
>>> partition /vicepa of vorab
>>> # find /vicepa /vicepa
>>> /vicepa/lost+found
>>> /vicepa/Lock
>>> /vicepa/Lock/vicepa
>>> /vicepa/AFSIDat
>>> /vicepa/AFSIDat/README
>>> /vicepa/AFSIDat/8=
>>> /vicepa/AFSIDat/8=/8=++U
>>> /vicepa/AFSIDat/8=/8=++U/special
>>> /vicepa/AFSIDat/8=/8=++U/special/zzzz5
>>> /vicepa/AFSIDat/8=/8=++U/special/zzzz9
>>> /vicepa/AFSIDat/8=/8=++U/special/zzzzD
>>> /vicepa/AFSIDat/8=/8=++U/special/zzzzP
>>> /vicepa/AFSIDat/8=/8=++U/+
>>> /vicepa/AFSIDat/8=/8=++U/+/+
>>> /vicepa/AFSIDat/8=/8=++U/+/+/=
>>> /vicepa/V0536870986.vl
>>> # df /vicepa Filesystem 512-blocks Free %Used Iused
>>> %Iused Mounted on
>>> /dev/lv02 262144 253720 4% 32 1% /vicepa
>>> # fs mkm /afs/.t/test test
>>> # ls -ld /afs/.t/test
>>> drwxrwxrwx 2 root system 2048 Aug 27 17:36 /afs/.t/test
>>> # bos salvage vorab /vicepa
>>> Note: Operation is performed on cell test.zurich.ibm.com
>>> bos: shutting down fs.
>>> Starting salvage.
>>> bos: salvage completed
>>> bos: restarting fs.
>>> # bos getlog vorab SalvageLog
>>> Fetching log file 'SalvageLog'...
>>> Note: Operation is performed on cell test.zurich.ibm.com
>>> @(#) OpenAFS 1.2.6 built 2002-08-21 08/27/2002 17:37:31 STARTING AFS
>>> SALVAGER 2.4 (/usr/afs/bin/salvager -f /vicepa)
>>> 08/27/2002 17:37:31 SALVAGING FILE SYSTEM PARTITION /vicepa
>>> (device=vicepa)
>>> 08/27/2002 17:37:31 ***Forced salvage of all volumes on this
>>> partition***
>>> 08/27/2002 17:37:32 SALVAGING VOLUME 536870986.
>>> 08/27/2002 17:37:32 test (536870986) not updated (created 08/27/2002
>>> 17:36)
>>> 08/27/2002 17:37:32 Vnode 1: vnode.unique, 1, does not match inode
>>> unique, 0; fixed, but status will be wrong
>>> 08/27/2002 17:37:32 Vnode 1: link count incorrect (was 2, now 0)
>>> 08/27/2002 17:37:32 Salvaged test (536870986): 1 files, 2 blocks
>>> 08/27/2002 17:37:32 SALVAGING OF PARTITION /vicepa COMPLETED
>>> # bos getlog vorab FileLog
>>> Fetching log file 'FileLog'...
>>> Note: Operation is performed on cell test.zurich.ibm.com
>>> Tue Aug 27 17:37:36 2002 File server starting
>>> Tue Aug 27 17:37:37 2002 afs_krb_get_lrealm failed, using
>>> test.zurich.ibm.com.
>>> Tue Aug 27 17:37:37 2002 Partition /vicepa: attached 1 volumes; 0
>>> volumes not attached
>>> Tue Aug 27 17:37:37 2002 Getting FileServer name...
>>> Tue Aug 27 17:37:37 2002 FileServer host name is 'vorab.zurich.ibm.com'
>>> Tue Aug 27 17:37:37 2002 Getting FileServer address...
>>> Tue Aug 27 17:37:37 2002 FileServer vorab.zurich.ibm.com has address
>>> 9.4.3.252 (0x90403fc or 0x90403fc in host byte order)
>>> Tue Aug 27 17:37:37 2002 File Server started Tue Aug 27 17:37:37 2002
>>> # find /vicepa /vicepa
>>> /vicepa/lost+found
>>> /vicepa/Lock
>>> /vicepa/Lock/vicepa
>>> /vicepa/AFSIDat
>>> /vicepa/AFSIDat/README
>>> /vicepa/AFSIDat/8=
>>> /vicepa/AFSIDat/8=/8=++U
>>> /vicepa/AFSIDat/8=/8=++U/special
>>> /vicepa/AFSIDat/8=/8=++U/special/zzzz5
>>> /vicepa/AFSIDat/8=/8=++U/special/zzzz9
>>> /vicepa/AFSIDat/8=/8=++U/special/zzzzD
>>> /vicepa/AFSIDat/8=/8=++U/special/zzzzP
>>> /vicepa/AFSIDat/8=/8=++U/+
>>> /vicepa/AFSIDat/8=/8=++U/+/+
>>> /vicepa/AFSIDat/8=/8=++U/+/+/=
>>> /vicepa/V0536870986.vl
>>> # ls -ld /afs/.t/test ls: 0653-341 The file /afs/.t/test does not
>>> exist.
>>> # df /vicepa Filesystem 512-blocks Free %Used
>>> Iused %Iused Mounted on
>>> /dev/lv02 262144 253720 4% 32 1% /vicepa
>>>
>>
>>
>
>
--
-----------------------------------------------------------------
Hartmut Reuter e-mail reuter@rzg.mpg.de
phone +49-89-3299-1328
RZG (Rechenzentrum Garching) fax +49-89-3299-1301
Computing Center of the Max-Planck-Gesellschaft (MPG) and the
Institut fuer Plasmaphysik (IPP)
-----------------------------------------------------------------