smbsh segfault

[Stephen Langasek]

On Tue, 28 Sep 1999 Keith G. Murphy wrote:

> Steve Litt wrote:

> > smbsh doesn't work with glibc-2.1 on Linux systems. That includes RedHat
> > 6.0 and many other recent Linux distributions. It is very hard to fix this
> > as the glibc maintainers have deliberately removed the necessary hooks for
> > smbsh to work. They don't like the idea of user space filesystems.

> > The only thing we can suggest right now is to use smbfs instead.

> Any chance we could ever see some *cooperation* between the smbfs people
> (person?) and the Samba people?  There seems to be some bitchiness, NIH
> syndrome, or something going on.  The Samba FAQ even mentions smbfs as a
> "constant cause of complaint".  This seems inappropriate, seeing as how
> it works pretty well for most of us, and Samba doesn't really offer
> anything comparable, far as I can tell.  As we have seen, smbsh ain't
> there yet.  I also would much rather use a real file system than a shell
> replacement, and I suspect others feel the same.

Someone on the Samba team (Andrew, I believe?) has lately been taking an
active role in keeping smbmount and the smbfs stuff working smoothly, and
he's done a great job--although it'll be nice when I don't have to keep
relearning the smbmount syntax. :) Part of the trouble in the past was that
people came to the Samba team with smbfs problems, and not only did the
Samba team not have the answers (since they weren't responsible for the
smbfs implementation), they had no one they could refer questions to!
(I.e., the Linux smbfs code was without a maintainer for some time.)

Frankly, I would much rather be using smbsh than smbmount.  I don't want to
have to give users /any/ root access, even through a specialized suid-root
program like smbmnt.  One less suid-root program on the system is one less
source of potential exploits I have to worry about.  (Although I'm happy to
say that, as a rule, the Samba team are very security-conscious wrt their
code.)

smbsh also offers the promise of integrating workgroup browsing into the
virtual fs.  smbmount certainly can't do that.

If you'd rather use kernel fs support than user-space virtual fs support,
that's fine--you have that option, at least under Linux.  No one's stepped
forward yet to code a smbfs driver for other OSes, more's the pity.  But it
would be nice to have the user-space driver too, wouldn't it?  I hadn't
realized the glibc developers *deliberately* sabotaged smbsh in glibc 2.1.
I must say, I'm more than a little put out now that I know...

-Steve Langasek
postmodern programmer