Alexander Viro wrote:
>
> On Fri, 23 Mar 2001, Dave Kleikamp wrote:
>
> > Are you okay with JFS using a few ioctl's in the utilities and
> > continuing the discussion of whether more general-purpose metadata
> > access should occur in procfs or under your proposal?
>
> OK, one immediate problem with ioctls on directories: unions.
> Think what happens when you union-mount JFS somewhere. And
> mountpoint also is on JFS. Where should ioctl() go?
AFAIC, I wouldn't object to calling this a limitation and letting the
ioctl fail. In fact, the utilities could detect a union-mount and fail
immediately. Although, someone, someday, will ask for it to work, won't
they?
> Other that that (and general ugliness of ioctls) - no problems. I
> really think that trick I've described would be cleaner, but that's
> a separate story. It's not like we had a moratorium on new ioctls,
> after all and JFS wouldn't be the first fs to do something like that.
> I don't think that it's a good idea, but the worst thing that can
> happen is inconvenience for union-mount setups. When union-mount
> gets to testable stage, that is - it's not like it was going into the
> tree before 2.5, anyway.
>
> procfs use for per-filesystem stuff is a different issue - that is just
> asking for bad breakage.
I won't argue this point without learning more about it. So far I've
only played with procfs to peek at some global data.
--
David Kleikamp
IBM Linux Technology Center
-
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to [EMAIL PROTECTED]
