nfs bug & df: Can I lock up my kernel and overflow this buffer?
Erik Trulsson
ertr1013 at student.uu.se
Tue May 10 16:10:44 PDT 2005
On Tue, May 10, 2005 at 05:16:16PM -0500, Billy Newsom wrote:
> Jonathan Noack wrote:
> >> Anyone tried that sort of stuff in fstab? I'm a little skeptical.
> >
> > I use "that sort of stuff" and have for a long time. Here's one of my
> > fstab lines:
> >
> > optimator:/usr/home /usr/home nfs rw,-3,-T,-r=32768,-w=32768 0 0
> >
> > It's obvious you don't believe me but why are you unwilling to try it
> > yourself?
>
> Well, because this fails to work on the commandline:
>
> #mount -o -s -x 2 -T dell:/nfs /dellbak
>
> I tried tons of different ways, never could get mount to do that, so I
> gave up on fstabbing options.
Note that mount(8) (as well as mount_nfs(8)) says about the -o flag
that "Options are specified with a -o flag followed by a comma
separated string of options."
If you try writing it as the manpage says, i.e. like
mount -o -s,-x=2,-T dell:/nfs /dellbak
it should work fine.
>
> Since the above mount command wouldn't even work, I figured I could
> forget about putting those same options (which mount calls illegal) in
> the fstab file. That's where the man pages only go so far. Without the
> examples you give, I was pretty sure that it was pointless to get fstab
> options to do what mount wouldn't.
>
> What it boils down to is that mount is fine with these options in fstab,
> but barfs when doing them on the commandline. That was so
> couter-intuitive, I went around it for the sake of getting things done.
It works fine on the commandline, you just got the syntax for the
commandline wrong.
>
> FreeBSD man pages are nice and all, but without a textbook siting by
> with some examples, it can be difficult. I learned Unix pretty much ad
> hoc, so I find that examples (such as you gave) are worth much more than
> man pages now that I know most of the basics.
Man pages often have some examples, but learning things only from
manpages can indeed be a bit difficult.
>
> Thanks.
>
> But what I did discover is that if I mount the same nfs resource
> multiple times, I get multiple, identical mounts (using fstab options,
> or commandline, either one). I have to umount each one serially. How
> is this a feature?
It is a feature in that the system does exactly what you asked it to
do. This is usually less painful than systems that try to guess what
you actually meant.
> What good does it do me if I mount the same nfs
> drive to the same place n times?
None that I can think of, which doesn't mean that there is no use for
it.
> Won't that eventually cause a deadlock
> as n increases beyond a few hundred or thousand? -- especially when the
> NFS server goes down?
Probably not a deadlock, but possible a resource starvation - so don't
do that.
> Shouldn't the second and subsequent mounts either
> fail or not be attempted due to a sanity check?
>
> #mount /usr
> mount: /dev/ad0s1f: Device busy
> Exit 1
>
> That seems reasonable for /usr. But as I stated before, NFS resources
> nevere apparently become "busy", and there is no sanity check to prevent
> mulitiple simultaneous mounts of identical file systems on identical
> file trees.
--
<Insert your favourite quote here.>
Erik Trulsson
ertr1013 at student.uu.se
More information about the freebsd-stable
mailing list