Re: GPF in shm_lock ipc

[Davidlohr Bueso]

On Mon, 12 Oct 2015, Kirill A. Shutemov wrote:

> On Mon, Oct 12, 2015 at 10:49:45AM -0700, Davidlohr Bueso wrote:
> > diff --git a/ipc/shm.c b/ipc/shm.c
> > index 4178727..9615f19 100644
> > --- a/ipc/shm.c
> > +++ b/ipc/shm.c
> > @@ -385,9 +385,25 @@ static struct mempolicy *shm_get_policy(struct vm_area_struct *vma,
> >  static int shm_mmap(struct file *file, struct vm_area_struct *vma)
> >  {
> > -	struct shm_file_data *sfd = shm_file_data(file);
> > +	struct file *vma_file = vma->vm_file;
> > +	struct shm_file_data *sfd = shm_file_data(vma_file);
> > +	struct ipc_ids *ids = &shm_ids(sfd->ns);
> > +	struct kern_ipc_perm *shp;
> >  	int ret;
> > +	rcu_read_lock();
> > +	shp = ipc_obtain_object_check(ids, sfd->id);
> > +	if (IS_ERR(shp)) {
> > +		ret = -EINVAL;
> > +		goto err;
> > +	}
> > +
> > +	if (!ipc_valid_object(shp)) {
> > +		ret = -EIDRM;
> > +		goto err;
> > +	}
> > +	rcu_read_unlock();
> > +
>
> Hm. Isn't it racy? What prevents IPC_RMID from happening after this point?

Nothing, but that is later caught by shm_open() doing similar checks. We
basically end up doing a check between ->mmap() calls, which is fair imho.
Note that this can occur anywhere in ipc as IPC_RMID is a user request/cmd,
and we try to respect it -- thus you can argue this race anywhere, which is
why we have EIDRM/EINVL. Ultimately the user should not be doing such hacks
_anyway_. So I'm not really concerned about it.

Another similar alternative would be perhaps to make shm_lock() return an
error, and thus propagate that error to mmap return. That way we would have
a silent way out of the warning scenario (afterward we cannot race as we
hold the ipc object lock). However, the users would now have to take this
into account...

     [validity check lockless]
     ->mmap()
     [validity check lock]

> Shouldn't we bump shm_nattch here? Or some other refcount?

At least not shm_nattach, as that would acknowledge a new attachment after
a valid IPC_RMID. But the problem is also with how we check for marked for
deletion segments -- ipc_valid_object() checking the deleted flag. As such,
we always rely on explicitly checking against the deleted flag.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/