[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: NCPFS, anyone?



On Fri, Sep 06, 2002 at 11:59:55AM +0200, Petr Vandrovec wrote:
> On  5 Sep 02 at 19:45, Roy Bixler wrote:
> > On Thu, Sep 05, 2002 at 05:53:26PM -0500, Roy Bixler wrote:
> > > Has anyone gotten NCPFS mounts to work in Woody?  Actually, I have but
> > > only on my old '386 box.  However, on the Sparc, I get this:
> > > 
> > > # ncpmount -S server -U rcb -A server /mnt
> > > Logging into SERVER as RCB
> > > Password: ********
> > > ncpmount: Invalid argument in mount(2)
> > > 
> > > A 'dmesg' command shows the following:
> > > ncp_read_super: kernel requires mount version 3
> 
> It should work, but only if you are using sparc binary on sparc kernel,
> or sparc64 binary on sparc64 kernel, otherwise sizeof(int), sizeof(long),
> sizeof(__kernel_uid_t) and other may differ between kernel and user.

That's probably it - the machine has an Ultrasparc IIe processor but,
as far as I understand about the Debian Woody distribution, the
binaries are all built in 32-bit mode.

> I have 
> patch to ncpfs & kernel (2.5.x) to pass mount options through ASCII string, 
> if you are interested.
> 
> But before you'll try that, try first
> "ncpmount -S server -U rcb -A server -3 /mnt" and
> "ncpmount -S server -U rcb -A server -4 /mnt". 
> 
> 2.4.x kernels should use mount version 4 (which allows for 32bit uid/gid), 
> but if you'll look at arch/sparc64/kernel/sys_sparc32.c, 
> do_ncp_super_data_conv, you'll find that this function (1) does not check
> version at all, (2) works only with -3 version, and (3) I do not see where 
> this function actually copies unchanged members: version, ncp_fd, time_out, 
> retry_count and flags...

The only thing that changes if I execute those two commands is, with
the former, I no longer get the warning about "ncp_read_super: kernel
requires mount version 3".  Otherwise, they both complain about
"ncpmount: Invalid argument in mount(2)".

> But if you'll rebuild ncpmount as a 64bit binary, 
> problems should disappear without need to fix kernel.

I used gcc-3.0 to build a 64-bit 'ncpmount' binary and I get a
'Segmentation fault' when I try to execute it.

I would be willing to try kernel patches, especially in a 2.4.x
kernel.  I'll even experiment with 2.5 kernel patches if I can get one
of those to build and it is stable enough.

Thanks,
Roy Bixler
rcb@ucp.uchicago.edu



Reply to: