[openbeos] Re: FS question
- From: "Axel Dörfler" <axeld@xxxxxxxxxxxxxxxx>
- To: openbeos@xxxxxxxxxxxxx
- Date: Wed, 16 Jan 2002 15:24:54 +0100 (MET)
> The problem is to make the StorageKit on other FS than BeFS ... FAT..
> ReiserFS..
> ie, how to solve the problem of node_refs and entry_refs
> My problem is : how can I initialize eg.: a BDirectory with
> entry_ref? get_ref_for_path can tranlate from filename/path to
inode
and device number with stat(). it's easy. But I didn't find any reverse
way to make an lookup/mapping from dev.inode number to
> path ( or fildescriptor is would be suitable)
Citing from the sample source code of the iso9660 file system add-on:
/* For each item on the disk (directory, file, etc), your filesystem
should allocate a vnode struct and
pass it back to the kernel when fs_read_vnode is called. This
struct is then passed back in to
your file system by functions that reference an item on the disk.
You'll need to be able to
create a vnode from a vnode id, either by hashing the id number or
encoding the information needed
to create the vnode in the vnode id itself. Vnode ids are assigned
by your file system when the
filesystem walk function is called. For this driver, the block
number is encoded in the upper bits
of the vnode id, and the offset within the block in the lower,
allowing me to just read the info
to fill in the vnode struct from the disk. When the kernel is done
with a vnode, it will call
fs_write_vnode (somewhat of a misnomer) where you should deallocate
the struct.
*/
The file system itself handles this translation - so you may introduce
a vnode layer which is understands linux' file system model (I haven't
looked at that one, so I don't know the differences).
Adios...
Axel.
Other related posts:
