Re: Files >4GB in XFS realtime partition

[James Chapman <jchapman@xxxxxxxxxxx>]

Hi Nathan,

Eric Sandeen asked me to reproduce this on an x86 but I've been unable 
to find time to do so yet. I'll get back to you as soon as I can. I'll 
try to test with a 2.4.29 kernel.

/james

Nathan Scott wrote:

> On Wed, May 18, 2005 at 03:10:08PM +0100, James Chapman wrote:
>
> > I'm seeing corrupt data when trying to read/write files >4GB in a 
> > realtime partition. Are there any known bugs in this area? Corruption 
> > only occurs when the file grows larger than 4GB. No problems are seen 
> > when using non-realtime XFS partitions.
> >
> > Kernel: 2.4.29
> > Arch:   MIPS
>
>
> Hmm, I can't reproduce this on i386 with a current (CVS) kernel,
> both buffered and direct realtime IO looks OK at offsets larger
> than 4GB.  Does this sequence fail for you too?...
>
> [root@bruce xfstests]# xfs_io -Rf /mnt/xfs0/realtime
> xfs_io> pwrite 5g 4k
> wrote 4096/4096 bytes at offset 5368709120
> 4 KiB, 1 ops; 0.0000 sec (15.024 MiB/sec and 3846.1538 ops/sec)
> xfs_io> quit
>
> [root@bruce xfstests]# od -x /mnt/xfs0/realtime 
> 0000000 0000 0000 0000 0000 0000 0000 0000 0000
> *
> 50000000000 cdcd cdcd cdcd cdcd cdcd cdcd cdcd cdcd
> *
> 50000010000
>
> bruce xfs-cmds/xfstests# xfs_io -Rdf /mnt/xfs0/realtime-direct
> xfs_io> pwrite 5g 4k
> wrote 4096/4096 bytes at offset 5368709120
> 4 KiB, 1 ops; 0.0000 sec (208.965 KiB/sec and 52.2411 ops/sec)
> xfs_io> q
> bruce xfs-cmds/xfstests# od -x /mnt/xfs0/realtime-direct
> 0000000 0000 0000 0000 0000 0000 0000 0000 0000
> *
> 50000000000 cdcd cdcd cdcd cdcd cdcd cdcd cdcd cdcd
> *
> 50000010000
>
> cheers.