RAID-6 corruption

RAID-6 corruption

[Jim Paris]

[-- Attachment #1: Type: text/plain, Size: 689 bytes --]

Hello,

(I've posted some messages about this to linux-raid, but haven't
gotten much of a response; this time I'm cc'ing to linux-kernel)

I am seeing reproducible corruption with software RAID-6 on an array
that has been created with missing disks.  The attached script
demonstrates this easily through loopback devices, but I see the
problem just the same on real drives.

I do not see the problem with full RAID-6 arrays (no missing disks),
nor do I see it on RAID-5 arrays on the same system.  I haven't seen
the corruption when writing directly to the md device, but I haven't
tested that very much.

Tested on 2.6.6, 2.6.7, 2.6.8-rc2, with mdadm 1.5.0 and 1.6.0.
See attached.

-jim

[-- Attachment #2: r6test.sh --]
[-- Type: application/x-sh, Size: 965 bytes --]

Re: RAID-6 corruption

[H. Peter Anvin]

Followup to:  <20040729212046.GA27332@jim.sh>
By author:    Jim Paris <jim@jtan.com>
In newsgroup: linux.dev.raid
> 
> Hello,
> 
> (I've posted some messages about this to linux-raid, but haven't
> gotten much of a response; this time I'm cc'ing to linux-kernel)
> 

I haven't seen any of those messages, so this is the first case happening.

> I am seeing reproducible corruption with software RAID-6 on an array
> that has been created with missing disks.  The attached script
> demonstrates this easily through loopback devices, but I see the
> problem just the same on real drives.
> 
> I do not see the problem with full RAID-6 arrays (no missing disks),
> nor do I see it on RAID-5 arrays on the same system.  I haven't seen
> the corruption when writing directly to the md device, but I haven't
> tested that very much.
> 
> Tested on 2.6.6, 2.6.7, 2.6.8-rc2, with mdadm 1.5.0 and 1.6.0.
> See attached.

I will look at it.

	-hpa