xfs
[Top] [All Lists]

Re: Corruption of in-memory data detected.

To: Bryan-TheBS-Smith <b.j.smith@xxxxxxxx>
Subject: Re: Corruption of in-memory data detected.
From: Steve Lord <lord@xxxxxxx>
Date: Thu, 25 Oct 2001 11:36:17 -0500
Cc: Utz Lehmann <ulehmann@xxxxxxxxxxxxxx>, Steve Lord <lord@xxxxxxx>, "Martin K. Petersen" <mkp@xxxxxxx>, Marc Schmitt <schmitt@xxxxxxxxxxx>, Eric Sandeen <sandeen@xxxxxxx>, linux-xfs@xxxxxxxxxxx, florin@xxxxxxx
In-reply-to: Message from Bryan-TheBS-Smith <b.j.smith@ieee.org> of "Thu, 25 Oct 2001 12:32:27 EDT." <3BD83E9B.E9EEFD4F@ieee.org>
Sender: owner-linux-xfs@xxxxxxxxxxx
> Utz Lehmann wrote:
> > ok, i'm stupid. He has only 1.2TB and the default blocksize is
> > 1k.  The test doesn't work until he can set his md device to
> > 512 byte blocksize from userspace.
> 
> I don't know if this has any relevance, but don't the underlying
> 3Ware cards use a much bigger blocksize???  If this isn't an issue
> from a compatibility standpoint, how about a performance one?
> 
> Or am I totally missing the mark here?
> 
> Are there any other limitations with using such small blocksizes? 
> Even at 512bytes, we should be able to get 2TB (4Gblocks x 512
> bytes), right?
> 
> -- TheBS
> 

Disk addresses always get converted to 512 byte sectors in the block
layer anyway - the 2 Tbyte limit is pretty firmly wedged in the block
layer right now.

Steve

p.s. I am pretty sure I have a fix - and it is not in the kernel! mkfs 
     with the automatic inode sizing is broken above the 1 Tbyte
     boundary, fix coming shortly.





<Prev in Thread] Current Thread [Next in Thread>