Mailing List Archive


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [tlug] xfs, jfs, reiser



B0Ti wrote:

>On Tue, 07 May 2002 12:38:06 +0900
>Bruno Raoult <br@example.com> wrote:
>
>
>  
>
>>I have comparable recovery time with reiserfs (from 1 to 4 seconds). But > I must
>>admin that I have got 1 or 2 time strange behavior.
>>
>>For instance, one day, after a crash, a file in my .mozilla was not 
>>consistent: trying
>>to access it caused the reading program to be "killed", such as:
>>    % cat thefile
>>    killed.
>>    %
>>
>>I *never* had this kind of problem with ext2 (or ufs on Solaris). 
>>Sometimes I lost
>>data, but after a fsck the file system was always consistent.
>>    
>>
>
>I don't know if your problem is related but sometimes garbage shows up in files after a crash with reiserfs. AFAIK it is due to the fact that reiserfs is only journaling metadata only (The filesystem itself will be consistent after recovery, but the data in the files is not guaranteed to be).
>Ext3 is considered to be much safer, and it can also journal data. 
>A while ago there was an interview with the ext3 fs lead developer on /. , he said that the above mentioned behaviour is considered as a security hole. (Imagine parts of /etc/shadow showing up accidentally in somebody else's file)
> 
>  
>
But even "cat" or "od" killed. And they does not care data. It means 
that some pointer was just going somewhere outside
the disk, or something like that. This is not what I call consistent 
metadata. I would accept to loose half of a file, but not
to loose more than that! For me, metadata also includes a good pointer 
to the real data, even if this one is corrupted. This
was not even the case...

Bruno.





Home | Main Index | Thread Index

Home Page Mailing List Linux and Japan TLUG Members Links