Misaligned raid0

Phillip Susi psusi at cfl.rr.com
Wed Dec 21 16:01:38 UTC 2005

How can it possibly be ok?  Here is a simple example using a stripe 
factor of 4 on two disks:

Disk 0: 0  8 16
         1  9 !!
         2 10
         3 11
Disk 1: 4 12 ??
         5 13
         6 14
         7 15

That is a stripe factor of 4 with a total length of 18 sectors on a pair 
of disks that are only 9 sectors long.  Logical sector 16 maps to the 
last sector on disk 0, but 17 also maps to disk 0, but beyond the end of 
the disk, rather than mapping to the last sector on disk 1.

This is why it seems impossible to me to have a stripe length that is 
not an even multiple of the stripe width.

By the way, it turned out that the filesystem does end before the final 
stripe and it isn't reiserfs that is accessing past the end of the 
device, it's some kind of probing that is going on in the initramfs. 
Maybe LVM or EVMS or something.  I stopped worrying about it in any case.

Heinz Mauelshagen wrote:
> Hrm, partial strides should be ok with device-mapper.
> Seems like reiser decided to allocate past EOD.
> Heinz

More information about the Ataraid-list mailing list