On Thu, Feb 20, 2003 at 03:52:42PM -0800, Konrad Schroder wrote:
> One of the things I'd like to do with LFS is to make the system calls into
> ioctl calls, which in theory should let LFS be loaded as an LKM.  No other
> filesystem (except fifofs and specfs) has VOP_IOCTL defined as anything
> other than a stub, though, and I find that even if it is defined the ioctl
> still returns with EPASSTHROUGH.
> 
> Would anyone object if I let ioctls to VREG and VDIR fall through to the
> VOP_IOCTL case?

That's fine with me, as Jason said, that functionality can be used
in other ways as well.

Btw, how do you plan to use VOP_IOCTL here? Most lfs systems calls
take an fsid, and if you want to end up in a newly created lfs_ioctl,
you'd need an actual LFS vnode from an open file on an LFS filesystem,
not the device itself.

- Frank