Commit ad133ba3dc283300e5b62b5b7211d2f39fbf6ee7

Authored by Oleg Nesterov
Committed by Ingo Molnar
1 parent 29d7b90c15

sched, signals: fix the racy usage of ->signal in account_group_xxx/run_posix_cpu_timers

Impact: fix potential NULL dereference

Contrary to ad474caca3e2a0550b7ce0706527ad5ab389a4d4 changelog, other
acct_group_xxx() helpers can be called after exit_notify() by timer tick.
Thanks to Roland for pointing out this. Somehow I missed this simple fact
when I read the original patch, and I am afraid I confused Frank during
the discussion. Sorry.

Fortunately, these helpers work with current, we can check ->exit_state
to ensure that ->signal can't go away under us.

Also, add the comment and compiler barrier to account_group_exec_runtime(),
to make sure we load ->signal only once.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Showing 2 changed files with 16 additions and 6 deletions Side-by-side Diff

kernel/posix-cpu-timers.c
... ... @@ -1308,9 +1308,10 @@
1308 1308 */
1309 1309 static inline int fastpath_timer_check(struct task_struct *tsk)
1310 1310 {
1311   - struct signal_struct *sig = tsk->signal;
  1311 + struct signal_struct *sig;
1312 1312  
1313   - if (unlikely(!sig))
  1313 + /* tsk == current, ensure it is safe to use ->signal/sighand */
  1314 + if (unlikely(tsk->exit_state))
1314 1315 return 0;
1315 1316  
1316 1317 if (!task_cputime_zero(&tsk->cputime_expires)) {
... ... @@ -1323,6 +1324,8 @@
1323 1324 if (task_cputime_expired(&task_sample, &tsk->cputime_expires))
1324 1325 return 1;
1325 1326 }
  1327 +
  1328 + sig = tsk->signal;
1326 1329 if (!task_cputime_zero(&sig->cputime_expires)) {
1327 1330 struct task_cputime group_sample;
1328 1331  
kernel/sched_stats.h
... ... @@ -298,9 +298,11 @@
298 298 {
299 299 struct signal_struct *sig;
300 300  
301   - sig = tsk->signal;
302   - if (unlikely(!sig))
  301 + /* tsk == current, ensure it is safe to use ->signal */
  302 + if (unlikely(tsk->exit_state))
303 303 return;
  304 +
  305 + sig = tsk->signal;
304 306 if (sig->cputime.totals) {
305 307 struct task_cputime *times;
306 308  
307 309  
... ... @@ -325,9 +327,11 @@
325 327 {
326 328 struct signal_struct *sig;
327 329  
328   - sig = tsk->signal;
329   - if (unlikely(!sig))
  330 + /* tsk == current, ensure it is safe to use ->signal */
  331 + if (unlikely(tsk->exit_state))
330 332 return;
  333 +
  334 + sig = tsk->signal;
331 335 if (sig->cputime.totals) {
332 336 struct task_cputime *times;
333 337  
334 338  
... ... @@ -353,8 +357,11 @@
353 357 struct signal_struct *sig;
354 358  
355 359 sig = tsk->signal;
  360 + /* see __exit_signal()->task_rq_unlock_wait() */
  361 + barrier();
356 362 if (unlikely(!sig))
357 363 return;
  364 +
358 365 if (sig->cputime.totals) {
359 366 struct task_cputime *times;
360 367