netdev - [PATCH net-next v4] tcp: Don't acquire inet_listen

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-next>] [day] [month] [year] [list]
Message-ID: <YgQOebeZ10eNx1W6@linutronix.de>
Date:   Wed, 9 Feb 2022 19:56:57 +0100
From:   Sebastian Andrzej Siewior <bigeasy@...utronix.de>
To:     netdev@...r.kernel.org
Cc:     Jakub Kicinski <kuba@...nel.org>,
        Kuniyuki Iwashima <kuniyu@...zon.co.jp>,
        eric.dumazet@...il.com, davem@...emloft.net, dsahern@...nel.org,
        efault@....de, tglx@...utronix.de, yoshfuji@...ux-ipv6.org,
        Martin KaFai Lau <kafai@...com>
Subject: [PATCH net-next v4] tcp: Don't acquire inet_listen_hashbucket::lock
 with disabled BH.

Commit
   9652dc2eb9e40 ("tcp: relax listening_hash operations")

removed the need to disable bottom half while acquiring
listening_hash.lock. There are still two callers left which disable
bottom half before the lock is acquired.

On PREEMPT_RT the softirqs are preemptible and local_bh_disable() acts
as a lock to ensure that resources, that are protected by disabling
bottom halves, remain protected.
This leads to a circular locking dependency if the lock acquired with
disabled bottom halves is also acquired with enabled bottom halves
followed by disabling bottom halves. This is the reverse locking order.
It has been observed with inet_listen_hashbucket::lock:

local_bh_disable() + spin_lock(&ilb->lock):
  inet_listen()
    inet_csk_listen_start()
      sk->sk_prot->hash() := inet_hash()
	local_bh_disable()
	__inet_hash()
	  spin_lock(&ilb->lock);
	    acquire(&ilb->lock);

Reverse order: spin_lock(&ilb2->lock) + local_bh_disable():
  tcp_seq_next()
    listening_get_next()
      spin_lock(&ilb2->lock);
	acquire(&ilb2->lock);

  tcp4_seq_show()
    get_tcp4_sock()
      sock_i_ino()
	read_lock_bh(&sk->sk_callback_lock);
	  acquire(softirq_ctrl)	// <---- whoops
	  acquire(&sk->sk_callback_lock)

Drop local_bh_disable() around __inet_hash() which acquires
listening_hash->lock. Split inet_unhash() and acquire the
listen_hashbucket lock without disabling bottom halves; the inet_ehash
lock with disabled bottom halves.

Reported-by: Mike Galbraith <efault@....de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@...utronix.de>
Link: https://lkml.kernel.org/r/12d6f9879a97cd56c09fb53dee343cbb14f7f1f7.camel@gmx.de
Link: https://lkml.kernel.org/r/X9CheYjuXWc75Spa@hirez.programming.kicks-ass.net
---
v3…v4:
   - Reposted targeting the net-next tree.

v2…v3:
   - Update commit description with ilb2 instead ilb (requested by
     Martin KaFai Lau).

v1…v2: 
   Reposted with fixes and net-tree as requested. Please keep in mind that
   this only effects the PREEMPT_RT preemption model and I'm posting this
   as part of the merging efforts. Therefore I didn't add the Fixes: tag
   and used net-next as I didn't expect any -stable backports (but then
   Greg sometimes backports RT-only patches since "it makes the life of
   some folks easier" as he puts it).

 net/ipv4/inet_hashtables.c  |   53 ++++++++++++++++++++++++++------------------
 net/ipv6/inet6_hashtables.c |    5 ----
 2 files changed, 33 insertions(+), 25 deletions(-)

--- a/net/ipv4/inet_hashtables.c
+++ b/net/ipv4/inet_hashtables.c
@@ -637,7 +637,9 @@ int __inet_hash(struct sock *sk, struct
 	int err = 0;
 
 	if (sk->sk_state != TCP_LISTEN) {
+		local_bh_disable();
 		inet_ehash_nolisten(sk, osk, NULL);
+		local_bh_enable();
 		return 0;
 	}
 	WARN_ON(!sk_unhashed(sk));
@@ -669,45 +671,54 @@ int inet_hash(struct sock *sk)
 {
 	int err = 0;
 
-	if (sk->sk_state != TCP_CLOSE) {
-		local_bh_disable();
+	if (sk->sk_state != TCP_CLOSE)
 		err = __inet_hash(sk, NULL);
-		local_bh_enable();
-	}
 
 	return err;
 }
 EXPORT_SYMBOL_GPL(inet_hash);
 
-void inet_unhash(struct sock *sk)
+static void __inet_unhash(struct sock *sk, struct inet_listen_hashbucket *ilb)
 {
-	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
-	struct inet_listen_hashbucket *ilb = NULL;
-	spinlock_t *lock;
-
 	if (sk_unhashed(sk))
 		return;
 
-	if (sk->sk_state == TCP_LISTEN) {
-		ilb = &hashinfo->listening_hash[inet_sk_listen_hashfn(sk)];
-		lock = &ilb->lock;
-	} else {
-		lock = inet_ehash_lockp(hashinfo, sk->sk_hash);
-	}
-	spin_lock_bh(lock);
-	if (sk_unhashed(sk))
-		goto unlock;
-
 	if (rcu_access_pointer(sk->sk_reuseport_cb))
 		reuseport_stop_listen_sock(sk);
 	if (ilb) {
+		struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
+
 		inet_unhash2(hashinfo, sk);
 		ilb->count--;
 	}
 	__sk_nulls_del_node_init_rcu(sk);
 	sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
-unlock:
-	spin_unlock_bh(lock);
+}
+
+void inet_unhash(struct sock *sk)
+{
+	struct inet_hashinfo *hashinfo = sk->sk_prot->h.hashinfo;
+
+	if (sk_unhashed(sk))
+		return;
+
+	if (sk->sk_state == TCP_LISTEN) {
+		struct inet_listen_hashbucket *ilb;
+
+		ilb = &hashinfo->listening_hash[inet_sk_listen_hashfn(sk)];
+		/* Don't disable bottom halves while acquiring the lock to
+		 * avoid circular locking dependency on PREEMPT_RT.
+		 */
+		spin_lock(&ilb->lock);
+		__inet_unhash(sk, ilb);
+		spin_unlock(&ilb->lock);
+	} else {
+		spinlock_t *lock = inet_ehash_lockp(hashinfo, sk->sk_hash);
+
+		spin_lock_bh(lock);
+		__inet_unhash(sk, NULL);
+		spin_unlock_bh(lock);
+	}
 }
 EXPORT_SYMBOL_GPL(inet_unhash);
 
--- a/net/ipv6/inet6_hashtables.c
+++ b/net/ipv6/inet6_hashtables.c
@@ -333,11 +333,8 @@ int inet6_hash(struct sock *sk)
 {
 	int err = 0;
 
-	if (sk->sk_state != TCP_CLOSE) {
-		local_bh_disable();
+	if (sk->sk_state != TCP_CLOSE)
 		err = __inet_hash(sk, NULL);
-		local_bh_enable();
-	}
 
 	return err;
 }