Hi,
> >>> result = nfs_sys_mount(mi, options);
> >
> > No, nfs_sys_mount() does not use mi->extra_opts at all, only the
> > binary options.
>
> This is the text-based code, which I wrote. nfs_sys_mount() passes an options string (NUL-terminated C string) to the kernel, not a binary object. That string contains all the FS-specific mount options specified by the user.
>
> But your patch makes that string empty, by my reading. I think this is incorrect.
Ok, I'm happy to go through this line-by-line.
static int nfs_sys_mount(struct nfsmount_info *mi, struct mount_options *opts)
{
char *options = NULL;
int result;
if (mi->fake)
return 1;
if (po_join(opts, &options) == PO_FAILED) { // HERE
errno = EIO;
return 0;
}
result = mount(mi->spec, mi->node, mi->type,
mi->flags & ~(MS_USER|MS_USERS), options);
...
}
nfs_sys_mount() constructs the text options _itself_ purely from the
opts (2nd) argument HERE -- po_join has opts as input and options as
output.
My patch only changes the first argument (mi). So, no functional change
within nfs_sys_mount() at all.
The functional change is that with my patch, mi->extra_opts is kept
unchanged unless the system call is successful. mi->extra_opts is
actually reused later throughout the mount program, because
nfsmount_string() has extra_opts as an input _and_ output argument,
and propagates mi->extra_opts into the extra_opts variable in main.c.
I have actually tested this patch extensively and it fixes the
problem.
Regards,
Wolfram.
--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
