Tom Vijlbrief wrote:
> Ofcourse, you're right. But currently my ext2 Linux filesystem
> is limited to 2 Gb.
>
> What filesystem are you using that support files bigger
> than 2Gb?
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.
The VFS/VM problems are fixed in the 2.4. kernels (and there
are patches available for 2.2. kernels), and for the syscall
interface, recent kernels and recent glibcs work together to
implement the LFS interface (read64, lseek64 etc).
So, you need either a 2.4. kernel or a patched 2.2. kernel,
and a recent glibc (I think 2.1.3 is OK for the patched 2.2.
kernels, for 2.4. you would need glibc 2.2). Plus, of course,
you have to either change your applications to use read64 etc,
or recompile them with the appropriate LFS defines (I forgot
the exact name ...) to have read automatically mapped to read64
etc.
> I wonder if e.g. using the "devfs" filesystem would lift the
> 2Gb limit of real partitions? Has someone devfs up and running and
> could he/she test if llseek works on the equivalent of /dev/hda
> for offsets larger than 2Gb?
This has certainly nothing to do with devfs (devfs only changes
the way devices are named, not how they are implemented).
But AFAIK llseek on *device nodes* (not regular files) is
supposed to seek beyond 2Gb anyway, without any LFS changes.
[Why else would you need llseek in the first place?]
Is this not correct?
Bye,
Ulrich
--
Dr. Ulrich Weigand
[EMAIL PROTECTED]
