Fix a race in tty_signal_sessleader() with unlocked read of s_leader.

Since we do not own the session lock, a parallel killjobc() might
reset s_leader to NULL after we checked it.  Read s_leader only once
and ensure that compiler is not allowed to reload.

While there, make access to t_session somewhat more pretty by using
local variable.

PR:	251915
Submitted by:	Jakub Piecuch <j.piecuch96@gmail.com>
MFC after:	1 week
This commit is contained in:
Konstantin Belousov 2020-12-17 19:51:39 +00:00
parent 0ebcbcc0ef
commit 551e205f6d
Notes: svn2git 2020-12-20 02:59:44 +00:00
svn path=/head/; revision=368735

View File

@ -1474,6 +1474,7 @@ void
tty_signal_sessleader(struct tty *tp, int sig)
{
struct proc *p;
struct session *s;
tty_assert_locked(tp);
MPASS(sig >= 1 && sig < NSIG);
@ -1482,8 +1483,14 @@ tty_signal_sessleader(struct tty *tp, int sig)
tp->t_flags &= ~TF_STOPPED;
tp->t_termios.c_lflag &= ~FLUSHO;
if (tp->t_session != NULL && tp->t_session->s_leader != NULL) {
p = tp->t_session->s_leader;
/*
* Load s_leader exactly once to avoid race where s_leader is
* set to NULL by a concurrent invocation of killjobc() by the
* session leader. Note that we are not holding t_session's
* lock for the read.
*/
if ((s = tp->t_session) != NULL &&
(p = atomic_load_ptr(&s->s_leader)) != NULL) {
PROC_LOCK(p);
kern_psignal(p, sig);
PROC_UNLOCK(p);