Received-SPF: pass (google.com: domain of linux-kernel+bounces-121962-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:40f1:3f00::1 as permitted sender) client-ip=2604:1380:40f1:3f00::1;
References: <00000000000090fe770614a1ab17@google.com>
 <0000000000007a208d0614a9a9e0@google.com>
User-agent: mu4e 1.6.10; emacs 29.2
From: Jakub Sitnicki <jakub@cloudflare.com>
To: syzbot <syzbot+d4066896495db380182e@syzkaller.appspotmail.com>
Cc: andrii@kernel.org, ast@kernel.org, bpf@vger.kernel.org,
 daniel@iogearbox.net, davem@davemloft.net, edumazet@google.com,
 john.fastabend@gmail.com, kuba@kernel.org, linux-kernel@vger.kernel.org,
 netdev@vger.kernel.org, pabeni@redhat.com, syzkaller-bugs@googlegroups.com
Subject: Re: [syzbot] [bpf?] [net?] possible deadlock in
 ahci_single_level_irq_intr
Date: Wed, 27 Mar 2024 21:04:14 +0100
In-reply-to: <0000000000007a208d0614a9a9e0@google.com>
Message-ID: <87le63bfuf.fsf@cloudflare.com>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain

#syz test: git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf.git master

diff --git a/net/core/sock_map.c b/net/core/sock_map.c
index 27d733c0f65e..3692f7256dd6 100644
--- a/net/core/sock_map.c
+++ b/net/core/sock_map.c
@@ -907,6 +907,7 @@ static void sock_hash_delete_from_link(struct bpf_map *map, struct sock *sk,
 	struct bpf_shtab_elem *elem_probe, *elem = link_raw;
 	struct bpf_shtab_bucket *bucket;
 
+	WARN_ON_ONCE(irqs_disabled());
 	WARN_ON_ONCE(!rcu_read_lock_held());
 	bucket = sock_hash_select_bucket(htab, elem->hash);
 
@@ -933,6 +934,10 @@ static long sock_hash_delete_elem(struct bpf_map *map, void *key)
 	struct bpf_shtab_elem *elem;
 	int ret = -ENOENT;
 
+	/* Can't run. We don't to play nice with hardirq-safe locks. */
+	if (irqs_disabled())
+		return -EOPNOTSUPP;
+
 	hash = sock_hash_bucket_hash(key, key_size);
 	bucket = sock_hash_select_bucket(htab, hash);
 
@@ -986,6 +991,7 @@ static int sock_hash_update_common(struct bpf_map *map, void *key,
 	struct sk_psock *psock;
 	int ret;
 
+	WARN_ON_ONCE(irqs_disabled());
 	WARN_ON_ONCE(!rcu_read_lock_held());
 	if (unlikely(flags > BPF_EXIST))
 		return -EINVAL;