| Message ID | 1275547159-30008-1-git-send-email-scott.lovenberg@gmail.com |
|---|---|
| State | New |
| Headers | show |
On Thu, 3 Jun 2010 02:39:19 -0400 Scott Lovenberg <scott.lovenberg@gmail.com> wrote: > The option parsing function now accepts all values for 'dir_mode' that are supported by the kernel side code. > > Signed-off-by: Scott Lovenberg <scott.lovenberg@gmail.com> > --- > mount.cifs.c | 2 +- > 1 files changed, 1 insertions(+), 1 deletions(-) > > diff --git a/mount.cifs.c b/mount.cifs.c > index 65754c0..21ce532 100644 > --- a/mount.cifs.c > +++ b/mount.cifs.c > @@ -812,7 +812,7 @@ static int parse_opt_token(const char *token) > return OPT_FILE_MODE; > if (strncmp(token, "dmask", 5) == 0) > return OPT_DMASK; > - if (strncmp(token, "dir_mode", 8) == 0) > + if (strncmp(token, "dir_mode", 4) == 0 || strncmp(token, "dirm", 4) == 0) ^^^^ Sigh. But I can confirm that this is similarly broken in the kernel so we have little choice but to live with it here. > return OPT_DIR_MODE; > if (strncmp(token, "nosuid", 6) == 0) > return OPT_NO_SUID; Committed...
> > > - if (strncmp(token, "dir_mode", 8) == 0) > > + if (strncmp(token, "dir_mode", 4) == 0 || strncmp(token, "dirm", 4) > == 0) > ^^^^ > Sigh. But I can confirm that this is similarly broken in the kernel > so we have little choice but to live with it here. > > That's exactly how I felt about this fix. Do you have any feelings on fixing this on both sides to accept only the full strings? It might cause regressions for people, so the kernel patch will probably be pulled by Redhat, Suse and Debian (and anyone else backporting kernel patches for long term releases) maintainers. It seems like the right thing to do, but if it's just going to cause extra work and be yanked either way... well, that's one step forward, two backwards.
On Mon, 7 Jun 2010 00:20:14 -0400 Scott Lovenberg <scott.lovenberg@gmail.com> wrote: > > > > > - if (strncmp(token, "dir_mode", 8) == 0) > > > + if (strncmp(token, "dir_mode", 4) == 0 || strncmp(token, "dirm", 4) > > == 0) > > ^^^^ > > Sigh. But I can confirm that this is similarly broken in the kernel > > so we have little choice but to live with it here. > > > > That's exactly how I felt about this fix. Do you have any feelings on > fixing this on both sides to accept only the full strings? It might cause > regressions for people, so the kernel patch will probably be pulled by > Redhat, Suse and Debian (and anyone else backporting kernel patches for long > term releases) maintainers. > It seems like the right thing to do, but if it's just going to cause extra > work and be yanked either way... well, that's one step forward, two > backwards. > > I think we're stuck living with it like this. The maintenance burden on this sort of thing is fairly low so it's not really a big deal. This is a good reason why we have to consider user-visible interfaces very carefully however.
diff --git a/mount.cifs.c b/mount.cifs.c index 65754c0..21ce532 100644 --- a/mount.cifs.c +++ b/mount.cifs.c @@ -812,7 +812,7 @@ static int parse_opt_token(const char *token) return OPT_FILE_MODE; if (strncmp(token, "dmask", 5) == 0) return OPT_DMASK; - if (strncmp(token, "dir_mode", 8) == 0) + if (strncmp(token, "dir_mode", 4) == 0 || strncmp(token, "dirm", 4) == 0) return OPT_DIR_MODE; if (strncmp(token, "nosuid", 6) == 0) return OPT_NO_SUID;
The option parsing function now accepts all values for 'dir_mode' that are supported by the kernel side code. Signed-off-by: Scott Lovenberg <scott.lovenberg@gmail.com> --- mount.cifs.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)