lkml.org 
[lkml]   [2006]   [Mar]   [30]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [RFC][PATCH 0/2]Extend ext3 filesystem limit from 8TB to 16TB
From
Date
On Thu, 2006-03-30 at 10:40 -0700, Andreas Dilger wrote:
> On Mar 29, 2006 17:38 -0800, Mingming Cao wrote:
> > Have verified these two patches on a 64 bit machine with 10TB ext3
> > filesystem, fsx runs fine for a few hours. Also testes on 32 bit machine
> > with <8TB ext3.
>
> Have you done tests _near_ 8TB with a 32-bit machine, even without these
> patches?
No I haven't. The >8TB right now is attached to a 64 bit machine, but we
should able to move it to a 32 bit machine.

> In particular, filling up the filesystem to be close to full
> so that we really depend on the > 2TB code to work properly?

I made a kernel patch to allow a file to specify which block group it
wants it's blocks to allocate from(using ioctl to set the goal
allocation block group). I set the goal block group falls to somewhere
>8TB, and did dd tests on that file. Verified this with debugfs, the
allocated block numbers are beyond 2**31.

Also before run fsx tests, created many directories (32768 at most:) and
verified one directory's inode is located in block group >8TB space. So
when we do fsx test on files under that directory, we are
creating/testing files >8TB.

BTW, do you think this ioctl is useful in general for other users? I
attached the patch here.

I also plan to hack the code of inode allocation to force all files's
inode is put in the block group >8TB, so that we could do a full
filesystem tests there.


> Also, in
> theory with these patches even a 32-bit machine could run > 8TB, right?
>
> There have been sporadic reports of failure for large ext3 filesystems,
> and some of them say that 32-bit systems fail and 64-bit systems work.
> There is a kernel bugzilla bug open for this, but it was never really
> identified what the source of the problem was.
>

Sure, I will verify that on my 32 bit machine with >8TB.

> Cheers, Andreas
> --
> Andreas Dilger
> Principal Software Engineer
> Cluster File Systems, Inc.
>

Thanks,

Mingming

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2006-03-30 21:19    [W:0.202 / U:1.400 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site