utrace_reap(), like other reporting loops, sets/clears utrace->reporting,
but this is not needed.
We already set ->ops = NULL, this means utrace_barrier() will never look
at reporting. get_utrace_lock() can not succeed, it must return -ESRCH
and utrace_barrier() just returns in this case.
But more importantly, I think utrace_reap() never needs to synchronize
with utrace_barrier(). Once we drop utrace->lock, any other caller which
takes this lock must see both ->exit_state and utrace->reap. This means
utrace_control() or utrace_set_events() which sets/clears REAP must not
succeed.
Signed-off-by: Oleg Nesterov <o...@redhat.com>
---
kernel/utrace.c | 9 ---------
1 file changed, 9 deletions(-)
--- __UTRACE/kernel/utrace.c~2_NO_REPORTING 2009-09-06 15:06:27.000000000
+0200
+++ __UTRACE/kernel/utrace.c 2009-09-06 15:09:59.000000000 +0200
@@ -429,19 +429,10 @@ restart:
if (!(flags & UTRACE_EVENT(REAP)))
continue;
- /*
- * This synchronizes with utrace_barrier(). Since we
- * need the utrace->lock here anyway (unlike the other
- * reporting loops), we don't need any memory barrier
- * as utrace_barrier() holds the lock.
- */
- utrace->reporting = engine;
spin_unlock(&utrace->lock);
(*ops->report_reap)(engine, target);
- utrace->reporting = NULL;
-
put_detached_list(&detached);
spin_lock(&utrace->lock);
