John Summerfield wrote:
> [EMAIL PROTECTED] said:
> > Ext2 is not limited to 2Gb. The problem you are seeing is a
> > combination of limitations inside the Linux VFS / VM layers, and the
> > 32-bit system call interface.
>
> Since Linux can use disks greater than 2 gbytes AND fdisk etc can manipulte
> partition information the CAN be spread all over them, I wonder about this. I
> presume dump/restore also can be used on disks greater than 2 Gbytes.
Well, accessing device nodes (i.e. *not* files on a filesystem) should
work above the 2Gb boundary; the buffer cache can handle large devices,
and the kernel provides the llseek system call.
It is only when accessing regular files, when the VFS and the page
cache come into play, that the 2Gb boundary hits. The most obvious
reason for this is that the page cache identifies pages by a pair
of (inode, file_offset), where the file_offset is a 32-bit signed
integer ... In 2.4, the file offset in *bytes* was replaced by
the offset in *pages*, raising the limit from 2Gb to 8Tb ...
For the user, this means that as long as only device nodes, not
regular files, are involved, *and* the tools used to access the
device are aware of the problem (i.e. use llseek()), then accessing
beyond 2Gb should work. As far as I'm aware of, this is the
case for tools typically used to access raw devices, e.g. fdisk
or dd, but not for 'normal' tools like cat ...
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
