ptrace: fix exit_ptrace() vs ptrace_traceme() race

Pointed out by Roland.  The bug was recently introduced by me in
"forget_original_parent: split out the un-ptrace part", commit
39c626ae47c469abdfd30c6e42eff884931380d6.

Since that patch we have a window after exit_ptrace() drops tasklist and
before forget_original_parent() takes it again.  In this window the child
can do ptrace(PTRACE_TRACEME) and nobody can untrace this child after
that.

Change ptrace_traceme() to not attach to the exiting ->real_parent.  We
don't report the error in this case, we pretend we attach right before
->real_parent calls exit_ptrace() which should untrace us anyway.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Roland McGrath <roland@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

diff --git a/kernel/ptrace.c b/kernel/ptrace.c
index 64191fa..dfcd83c 100644 (file)
--- a/kernel/ptrace.c
+++ b/kernel/ptrace.c
@@ -604,10 +604,11 @@ repeat:
                ret = security_ptrace_traceme(current->parent);
 
                /*
-                * Set the ptrace bit in the process ptrace flags.
-                * Then link us on our parent's ptraced list.
+                * Check PF_EXITING to ensure ->real_parent has not passed
+                * exit_ptrace(). Otherwise we don't report the error but
+                * pretend ->real_parent untraces us right after return.
                 */
-               if (!ret) {
+               if (!ret && !(current->real_parent->flags & PF_EXITING)) {
                        current->ptrace |= PT_PTRACED;
                        __ptrace_link(current, current->real_parent);
                }