On Thu, May 17, 2018 at 05:41:39PM +0200, Oleg Nesterov wrote:
> again, I don't understand this code yet, but
>
> On 05/17, Tycho Andersen wrote:
> >
> > +long seccomp_get_listener(struct task_struct *task,
> > + unsigned long filter_off)
> > +{
> > + struct seccomp_filter *filter;
> > + struct file *listener;
> > + int fd;
> > +
> > + filter = get_nth_filter(task, filter_off);
> > + if (IS_ERR(filter))
> > + return PTR_ERR(filter);
> > +
> > + fd = get_unused_fd_flags(O_RDWR);
> > + if (fd < 0) {
> > + __put_seccomp_filter(filter);
> > + return fd;
> > + }
> > +
> > + listener = init_listener(task, task->seccomp.filter);
> > + if (IS_ERR(listener)) {
> > + put_unused_fd(fd);
> > + return PTR_ERR(listener);
>
> __put_seccomp_filter() ?
Yes, I think you're right here.
> and since init_listener() does __get_seccomp_filter() on sucess, it is needed
> uncondtitionally?
I think there does need to be a __get_seccomp_filter() on success in
init_listener(), because it's paired with the __put_seccomp_filter in
seccomp_notify_release. The listener fd has a reference to the filter,
and that shouldn't go away until after the fd is freed.
Tycho
