FW: "attempt to access beyond end of device" on shared UDF
peter at helios.de
Fri Sep 3 07:11:51 GMT 2004
Am Donnerstag, 2. September 2004 21:06 schrieb Anthony Russello:
> > Hi,
> > I am trying to track down an issue in which writes to a UDF
> > volume that is shared via samba result in the following error:
> > <snip>
> > attempt to access beyond end of device
> > 0b:01: rw=1, want=1120061380, limit=34181760
> > </snip>
> > The 'want=' value is usually different, and can also be zero
> > in some instances.
> > The volume is an Iomega REV drive formatted using the UDF
> > filesystem. Here are the specifications for my test system:
> > kernel 2.4.26
> > applied the following patches:
> > - patches-be.dvd.ram-20040816 (from
> > http://iomrrdtools.sourceforge.net/)
> > samba version 3.0.5
> > The error occurs when I copy a folder with subfolders from a
> > Windows XP (both SP1 and SP2) workstation to a samba share
> > that is linked to the mount point for the Iomega REV drive.
> > It should be noted that I have not been able to reproduce
> > this issue when I connect to the same share using a linux
> > workstation (also running samba 3.0.5). Also, this error
> > does not occur when I am copying files and folders locally on
> > the workstation with the REV drive.
> > The fact that I cannot seem to reproduce the error by copying
> > files and folders locally has me thinking it might be related
> > to samba and XP writing to a samba share.
Then copy it several times into different subdirs locally.
This way you can provoke that the blocklayer will hit blocks
outside of the block device more likely.
I bet that this has nothing to do with Samba - Samba runs in user space
and does not choose the blocks for the filesystem.
The block numbers look like an overflow in the UDF or blockdevice layers.
More information about the samba-technical