Messages in this thread | | | From | (Matthias Urlichs) | Subject | Re: fmount system call | Date | 24 Jul 1997 12:30:03 +0200 |
| |
Jeremy Fitzhardinge <jeremy@zip.com.au> writes: > Theodore Y. Ts'o wrote: > > From: "Volker.Lendecke" <lendecke@math.uni-goettingen.de> > > > > As far as I can see this problem can be removed by a system call that > > accepts a file descriptor as a mount point. > > Yes, this is a good idea. > Seconded.
> I would agree, however, that mount shouldn't follow symlinks on the > mountpoint (I'm not sure if it does at the moment, and given the current > development kernels, I doubt if the kernel knows either). > The kernel shouldn't know or care how you get somewhere (with the exception of looped-back mounts which we don't have at the moment anyway), only that you can do what you want to do if you are there. Therefore, we need atomic operations. Since mount(8), as a setuid program which does separate per- missions checks, can't guarantee atomicity, we need an atomic mount call.
Simple, right? ;-)
| |