Re: [PATCH 2/3] Add extended attributes to ext2/3

[Theodore Ts'o]

On Wed, Oct 16, 2002 at 11:05:39AM -0600, Andreas Dilger wrote:
> Err, wouldn't that be a 2TB FILE limit, and not a FILESYSTEM limit?

Yes, sorry about that.

I just took a quick look at this, and the limit is unfortunately
fundamental to the VFS; the st_blocks in struct inode and struct kstat
is an unsigned long, which means that we're stuck with the 2TB limit
for *all* filesystems, even using the LFS API.  (Or rather, as you
point out, since no one apparently is doing any overflow checking, so
st_blocks will just be incorrect, even to callers using the stat64
call.)

A quick fix would be to make st_blocks in struct inode and struct
kstat take a __u64 instead of a unsigned long, but it will have a
performance impact.  What do folks think?

						- Ted