linux-kernel - [PATCH v4 4/4] locktorture: With nested locks, occasionally skip main lock

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20230221190238.21285-4-jstultz@google.com>
Date:   Tue, 21 Feb 2023 19:02:38 +0000
From:   John Stultz <jstultz@...gle.com>
To:     LKML <linux-kernel@...r.kernel.org>
Cc:     John Stultz <jstultz@...gle.com>,
        Davidlohr Bueso <dave@...olabs.net>,
        "Paul E. McKenney" <paulmck@...nel.org>,
        Josh Triplett <josh@...htriplett.org>,
        Joel Fernandes <joel@...lfernandes.org>,
        Juri Lelli <juri.lelli@...hat.com>,
        Valentin Schneider <vschneid@...hat.com>,
        Dietmar Eggemann <dietmar.eggemann@....com>,
        kernel-team@...roid.com, "Connor O'Brien" <connoro@...gle.com>
Subject: [PATCH v4 4/4] locktorture: With nested locks, occasionally skip main lock

If we're using nested locking to stress things, occasionally
skip taking the main lock, so that we can get some different
contention patterns between the writers (to hopefully get two
disjoint blocked trees)

Cc: Davidlohr Bueso <dave@...olabs.net>
Cc: "Paul E. McKenney" <paulmck@...nel.org>
Cc: Josh Triplett <josh@...htriplett.org>
Cc: Joel Fernandes <joel@...lfernandes.org>
Cc: Juri Lelli <juri.lelli@...hat.com>
Cc: Valentin Schneider <vschneid@...hat.com>
Cc: Dietmar Eggemann <dietmar.eggemann@....com>
Cc: kernel-team@...roid.com
Co-developed-by: Connor O'Brien <connoro@...gle.com>
Signed-off-by: Connor O'Brien <connoro@...gle.com>
Signed-off-by: John Stultz <jstultz@...gle.com>
---
v3:
* Minor rework for naming changes suggested by Davidlohr Bueso
* Added comment explaining why we skip the main lock, as
  suggested by Davidlohr Bueso
v4:
* Add co-developed tag
---
 kernel/locking/locktorture.c | 37 ++++++++++++++++++++++++++----------
 1 file changed, 27 insertions(+), 10 deletions(-)

diff --git a/kernel/locking/locktorture.c b/kernel/locking/locktorture.c
index fd8b2b007c26..ab896cc42b87 100644
--- a/kernel/locking/locktorture.c
+++ b/kernel/locking/locktorture.c
@@ -745,6 +745,7 @@ static int lock_torture_writer(void *arg)
 	int tid = lwsp - cxt.lwsa;
 	DEFINE_TORTURE_RANDOM(rand);
 	u32 lockset_mask;
+	bool skip_main_lock;
 
 	VERBOSE_TOROUT_STRING("lock_torture_writer task started");
 	set_user_nice(current, MAX_NICE);
@@ -754,21 +755,37 @@ static int lock_torture_writer(void *arg)
 			schedule_timeout_uninterruptible(1);
 
 		lockset_mask = torture_random(&rand);
+		/*
+		 * When using nested_locks, we want to occasionally
+		 * skip the main lock so we can avoid always serializing
+		 * the lock chains on that central lock. By skipping the
+		 * main lock occasionally, we can create different
+		 * contention patterns (allowing for multiple disjoint
+		 * blocked trees)
+		 */
+		skip_main_lock = (nested_locks &&
+				 !(torture_random(&rand) % 100));
+
 		cxt.cur_ops->task_boost(&rand);
 		if (cxt.cur_ops->nested_lock)
 			cxt.cur_ops->nested_lock(tid, lockset_mask);
-		cxt.cur_ops->writelock(tid);
-		if (WARN_ON_ONCE(lock_is_write_held))
-			lwsp->n_lock_fail++;
-		lock_is_write_held = true;
-		if (WARN_ON_ONCE(atomic_read(&lock_is_read_held)))
-			lwsp->n_lock_fail++; /* rare, but... */
 
-		lwsp->n_lock_acquired++;
+		if (!skip_main_lock) {
+			cxt.cur_ops->writelock(tid);
+			if (WARN_ON_ONCE(lock_is_write_held))
+				lwsp->n_lock_fail++;
+			lock_is_write_held = true;
+			if (WARN_ON_ONCE(atomic_read(&lock_is_read_held)))
+				lwsp->n_lock_fail++; /* rare, but... */
+
+			lwsp->n_lock_acquired++;
+		}
 		cxt.cur_ops->write_delay(&rand);
-		lock_is_write_held = false;
-		WRITE_ONCE(last_lock_release, jiffies);
-		cxt.cur_ops->writeunlock(tid);
+		if (!skip_main_lock) {
+			lock_is_write_held = false;
+			WRITE_ONCE(last_lock_release, jiffies);
+			cxt.cur_ops->writeunlock(tid);
+		}
 		if (cxt.cur_ops->nested_unlock)
 			cxt.cur_ops->nested_unlock(tid, lockset_mask);
 
-- 
2.39.2.637.g21b0678d19-goog