signals: check_kill_permission: check session under tasklist_lock
authorOleg Nesterov <oleg@tv-sign.ru>
Wed, 30 Apr 2008 07:53:01 +0000 (00:53 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Wed, 30 Apr 2008 15:29:37 +0000 (08:29 -0700)
This wasn't documented, but as Atsushi Tsuji pointed out
check_kill_permission() needs tasklist_lock for task_session_nr().  I missed
this fact when removed tasklist from the callers.

Change check_kill_permission() to take tasklist_lock for the SIGCONT case.
Re-order security checks so that we take tasklist_lock only if/when it is
actually needed.  This is a minimal fix for now, tasklist will be removed
later.

Also change the code to use task_session() instead of task_session_nr().

Also, remove the SIGCONT check from cap_task_kill(), it is bogus (and the
whole function is bogus.  Serge, Eric, why it is still alive?).

Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Acked-by: Atsushi Tsuji <a-tsuji@bk.jp.nec.com>
Cc: Roland McGrath <roland@redhat.com>
Cc: Serge Hallyn <serue@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
kernel/signal.c

index 251cc13720bdebdf6fc723bf5f2313f934b53fb1..24be82c0aae36ea636f8c346a67b7abcf7323590 100644 (file)
@@ -533,6 +533,7 @@ static int rm_from_queue(unsigned long mask, struct sigpending *s)
 static int check_kill_permission(int sig, struct siginfo *info,
                                 struct task_struct *t)
 {
+       struct pid *sid;
        int error;
 
        if (!valid_signal(sig))
@@ -545,11 +546,24 @@ static int check_kill_permission(int sig, struct siginfo *info,
        if (error)
                return error;
 
-       if (((sig != SIGCONT) || (task_session_nr(current) != task_session_nr(t)))
-           && (current->euid ^ t->suid) && (current->euid ^ t->uid)
-           && (current->uid ^ t->suid) && (current->uid ^ t->uid)
-           && !capable(CAP_KILL))
-               return -EPERM;
+       if ((current->euid ^ t->suid) && (current->euid ^ t->uid) &&
+           (current->uid  ^ t->suid) && (current->uid  ^ t->uid) &&
+           !capable(CAP_KILL)) {
+               switch (sig) {
+               case SIGCONT:
+                       read_lock(&tasklist_lock);
+                       sid = task_session(t);
+                       read_unlock(&tasklist_lock);
+                       /*
+                        * We don't return the error if sid == NULL. The
+                        * task was unhashed, the caller must notice this.
+                        */
+                       if (!sid || sid == task_session(current))
+                               break;
+               default:
+                       return -EPERM;
+               }
+       }
 
        return security_task_kill(t, info, sig, 0);
 }