> On Thu, 2 Dec 1999, Bruno Haible wrote:
>
> > 5) Use the proc filesystem. Implement a file /proc/ipc/shm/42 as being
> > equivalent to the shared memory segment with id 42.
> > File type: regular file
> > File size: the shm segment's size
> > File contents (for use by read, write, mmap): the shm segment's data
> > File owner/group: the shm segment's owner and group
> > truncate(): return -EINVAL
> >
> > Not only would this solve your "mmap of shared memory" problem, it would
> > become possible to view and edit shared memory using "cat", "hexdump" and
> > "vi". Benefits of the "everything is a file" philosophy.
>
> Don't do it in procfs. Make a separate filesystem and mount it on the
> empty directory in /proc, if you really need it (I'ld rather use some
> other location - even /dev/shm would be better). This filesystem will have
> _nothing_ with proc in terms of code. There is enough mess in procfs
> already. Keep this one separate.
I totally agree. After looking at the shm code I concluded that making
a filesystem out of it will cleanup the code a lot. There is much
duplication of code from the mm layer.
I am just working on this filesystem.
Greetings
Christoph
-- Christoph Rohland SAP AG LinuxLab Email: hans-christoph.rohland@sap.com- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/