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: <20250716144818.47650-10-boqun.feng@gmail.com>
Date: Wed, 16 Jul 2025 07:48:18 -0700
From: Boqun Feng <boqun.feng@...il.com>
To: Ingo Molnar <mingo@...nel.org>,
	Peter Zijlstra <peterz@...radead.org>
Cc: "Will Deacon" <will@...nel.org>,
	"Waiman Long" <longman@...hat.com>,
	Miguel Ojeda <ojeda@...nel.org>,
	Alice Ryhl <aliceryhl@...gle.com>,
	linux-kernel@...r.kernel.org,
	rust-for-linux@...r.kernel.org,
	Breno Leitao <leitao@...ian.org>,
	"Paul E. McKenney" <paulmck@...nel.org>,
	Erik Lundgren <elundgren@...a.com>,
	Eric Dumazet <edumazet@...gle.com>,
	Boqun Feng <boqun.feng@...il.com>
Subject: [PATCH 9/9] lockdep: Speed up lockdep_unregister_key() with expedited RCU synchronization

From: Breno Leitao <leitao@...ian.org>

lockdep_unregister_key() is called from critical code paths, including
sections where rtnl_lock() is held. For example, when replacing a qdisc
in a network device, network egress traffic is disabled while
__qdisc_destroy() is called for every network queue.

If lockdep is enabled, __qdisc_destroy() calls lockdep_unregister_key(),
which gets blocked waiting for synchronize_rcu() to complete.

For example, a simple tc command to replace a qdisc could take 13
seconds:

  # time /usr/sbin/tc qdisc replace dev eth0 root handle 0x1: mq
    real    0m13.195s
    user    0m0.001s
    sys     0m2.746s

During this time, network egress is completely frozen while waiting for
RCU synchronization.

Use synchronize_rcu_expedited() instead to minimize the impact on
critical operations like network connectivity changes.

This improves 10x the function call to tc, when replacing the qdisc for
a network card.

   # time /usr/sbin/tc qdisc replace dev eth0 root handle 0x1: mq
     real     0m1.789s
     user     0m0.000s
     sys      0m1.613s

[boqun: Fixed the comment and add more information for the temporary
workaround, and add TODO information for hazptr]

Reported-by: Erik Lundgren <elundgren@...a.com>
Signed-off-by: Breno Leitao <leitao@...ian.org>
Reviewed-by: Paul E. McKenney <paulmck@...nel.org>
Reviewed-by: Eric Dumazet <edumazet@...gle.com>
Signed-off-by: Boqun Feng <boqun.feng@...il.com>
Link: https://lore.kernel.org/r/20250321-lockdep-v1-1-78b732d195fb@debian.org
---
 kernel/locking/lockdep.c | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
index 0c941418a215..2d4c5bab5af8 100644
--- a/kernel/locking/lockdep.c
+++ b/kernel/locking/lockdep.c
@@ -6616,8 +6616,16 @@ void lockdep_unregister_key(struct lock_class_key *key)
 	if (need_callback)
 		call_rcu(&delayed_free.rcu_head, free_zapped_rcu);
 
-	/* Wait until is_dynamic_key() has finished accessing k->hash_entry. */
-	synchronize_rcu();
+	/*
+	 * Wait until is_dynamic_key() has finished accessing k->hash_entry.
+	 *
+	 * Some operations like __qdisc_destroy() will call this in a debug
+	 * kernel, and the network traffic is disabled while waiting, hence
+	 * the delay of the wait matters in debugging cases. Currently use a
+	 * synchronize_rcu_expedited() to speed up the wait at the cost of
+	 * system IPIs. TODO: Replace RCU with hazptr for this.
+	 */
+	synchronize_rcu_expedited();
 }
 EXPORT_SYMBOL_GPL(lockdep_unregister_key);
 
-- 
2.39.5 (Apple Git-154)


Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ