lkml.org 
[lkml]   [2012]   [Apr]   [19]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH] fs: print warning when mount flags was ignored
On Thu, 19 Apr 2012, Karel Zak wrote:

> On Thu, Apr 19, 2012 at 11:07:55AM +0200, Lukas Czerner wrote:
> > Some mount flags can conflict with each other so they can not be
> > handled together. Currently when conflicting flags are specified,
> > some of them are silently ignored putting user in believe that
> > they was handled correctly.
>
> Unfortunately, it's not so simple ;-)
>
> > - if (flags & MS_REMOUNT)
> > + if (flags & MS_REMOUNT) {
> > retval = do_remount(&path, flags & ~MS_REMOUNT, mnt_flags,
> > data_page);
> > - else if (flags & MS_BIND)
> > + flags &= ~MS_REMOUNT;
>
> This is incorrect, the flags may also include many others flags. For
> example MS_REMOUNT|MS_BIND|MS_RDONLY is valid (see do_remoun() code).

ah, right we can specify more flags to be consumed by do_remount and
others. Well, then we can just mask them all out like this:

flags &= ~(MS_REMOUNT | MS_BIND );

>
> And it's normal that for "mount -o remount" the mount command reads
> flags from mtab/fstab so it includes for example MS_RELATIME, ...

Yes, but MS_RELATIME is masked out from flags and it is used in
mnt_flags which are handled by remount. That makes me wonder why we even
have MS_RELATIME flag, since it is not used anywhere.

>
> > + } else if (flags & MS_BIND) {
> > retval = do_loopback(&path, dev_name, flags & MS_REC);
> > - else if (flags & (MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE))
> > + flags &= ~MS_BIND;
>
> what about MS_REC ?

Ok, we should put MS_REC into the mix since the do_remount does not use
it.

flags &= ~(MS_BIND | MS_REC);

>
> > + } else if (flags & (MS_SHARED | MS_PRIVATE |
> > + MS_SLAVE | MS_UNBINDABLE)) {
> > retval = do_change_type(&path, flags);
> > - else if (flags & MS_MOVE)
> > + flags &= ~(MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE);
>
> what about MS_REC ?

flags &= ~(MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE |
MS_REC);
>
> Note that do_change_type() already checks for unexpected flags and
> returns -EINVAL if more flags are specified.

I am not sure how this is related. Each of those function should
probably do that but do_change_type() is actually the only one
doing so.

>
> > + } else if (flags & MS_MOVE) {
> > retval = do_move_mount(&path, dev_name);
> > - else
> > + flags &= ~MS_MOVE;
> > + } else
> > retval = do_new_mount(&path, type_page, flags, mnt_flags,
> > dev_name, data_page);
> > +
> > + flags &= (MS_REMOUNT | MS_BIND | MS_SHARED | MS_PRIVATE |
> > + MS_SLAVE | MS_UNBINDABLE | MS_MOVE);

flags &= (MS_REMOUNT | MS_BIND | MS_SHARED | MS_PRIVATE |
MS_SLAVE | MS_UNBINDABLE | MS_MOVE | MS_REC);

> > +
> > + if (!retval && flags)
> > + printk(KERN_WARNING "%s(%u): (%s -> %s) Conflicting mount flags"
> > + " specified. These flags has been "
> > + "ignored: %#.8lx\n", __func__, current->pid,
> > + dev_name, dir_name, flags);

But then, we'll might ignoring mnt_flags silently since they are masked
out earlier. Well, maybe we could mas them out only do_remount() and
do_new_mount().

> Karel

Thanks!
-Lukas


\
 
 \ /
  Last update: 2012-04-19 13:51    [W:0.042 / U:0.756 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site