On Thu, Aug 24, 2017 at 03:20:41PM +0200, Oleg Nesterov wrote:
> On 08/23, Eric Biggers wrote:
> >
> > From: Eric Biggers <ebiggers@xxxxxxxxxx>
> >
> > Commit 7c051267931a ("mm, fork: make dup_mmap wait for mmap_sem for
> > write killable") made it possible to kill a forking task while it is
> > waiting to acquire its ->mmap_sem for write, in dup_mmap(). However, it
> > was overlooked that this introduced an new error path before a reference
> > is taken on the mm_struct's ->exe_file.
>
> Hmm. Unless I am totally confused, the same problem with mm->exol_area?
> I'll recheck....
I'm not sure what you mean by ->exol_area.
>
> > --- a/kernel/fork.c
> > +++ b/kernel/fork.c
> > @@ -806,6 +806,7 @@ static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p,
> > mm_init_cpumask(mm);
> > mm_init_aio(mm);
> > mm_init_owner(mm, p);
> > + RCU_INIT_POINTER(mm->exe_file, NULL);
>
> Can't we simply move
>
> RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm));
>
> from dup_mmap() here? Afaics this doesn't need mmap_sem.
>
Two problems, even assuming that get_mm_exe_file() doesn't require mmap_sem:
- If mm_alloc_pgd() or init_new_context() in mm_init() fails, mm_init() doesn't
do the full mmput(), so the file reference would not be dropped. So it would
need to be changed to drop the file reference too.
- The file would also be set when called from mm_alloc() which is used when
exec'ing a new task. *Maybe* it would be safe to do temporarily, but it's
pointless because ->exe_file will be set later by flush_old_exec().
Eric