On Wed, May 29, 2019 at 6:21 PM Oleg Nesterov <o...@redhat.com> wrote: > On 05/29, Jann Horn wrote: > > --- a/kernel/ptrace.c > > +++ b/kernel/ptrace.c > > @@ -324,6 +324,16 @@ static int __ptrace_may_access(struct task_struct > > *task, unsigned int mode) [...] > > mm = task->mm; > > while at it, could you also change this into mm = READ_ONCE(task->mm) ?
Actually, that shouldn't be necessary. The caller of __ptrace_may_access() holds the task_lock() on the task, and that should prevent concurrent updates of ->mm. If concurrent updates of ->mm *were* possible, we'd probably be in deep trouble here (and by that I mean use-after-free).
