Fine, but at least with EXT2, once the VFS is 64-bit, there
appears a surprise from rlimit data, which is defined as
"long", that is, signed 32-bits with 32-bit machines...
Perhaps that code should be changes to detect the magic
"unlimited" value (RLIM_INFINITY), and thus raise the limit
up to 4G, until the EXT2 can store larger i_size entries.
(fs/ext2/inode.c, look for "rlim_cur" variables.)
However there is no point in changeing it before the VFS
layer's i_size changes from off_t to size_t, or preferrably
to loff_t. (size_t is "unsigned long" at i386, which doesn't
help that much.)
> --
> Chip Salzenberg - a.k.a. - <chip@pobox.com>
/Matti Aarnio <matti.aarnio@tele.fi>