rds: tcp: must use spin_lock_irq* and not spin_lock_bh with rds_tcp_conn_lock

rds_tcp_connection allocation/free management has the potential to be
called from __rds_conn_create after IRQs have been disabled, so
spin_[un]lock_bh cannot be used with rds_tcp_conn_lock.

Bottom-halves that need to synchronize for critical sections protected
by rds_tcp_conn_lock should instead use rds_destroy_pending() correctly.

Reported-by: syzbot+c68e51bb5e699d3f8d91@syzkaller.appspotmail.com
Fixes: ebeeb1ad9b8a ("rds: tcp: use rds_destroy_pending() to synchronize
       netns/module teardown and rds connection/workq management")
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

diff --git a/net/rds/tcp.c b/net/rds/tcp.c
index eb04e7fa2467f0b3bcf01a4302fc11bc8bb09a65..08ea9cd5c2f65e617c8e21a22e86e8832d141735 100644 (file)
--- a/net/rds/tcp.c
+++ b/net/rds/tcp.c
@@ -272,13 +272,14 @@ static int rds_tcp_laddr_check(struct net *net, __be32 addr)
 static void rds_tcp_conn_free(void *arg)
 {
        struct rds_tcp_connection *tc = arg;
+       unsigned long flags;
 
        rdsdebug("freeing tc %p\n", tc);
 
-       spin_lock_bh(&rds_tcp_conn_lock);
+       spin_lock_irqsave(&rds_tcp_conn_lock, flags);
        if (!tc->t_tcp_node_detached)
                list_del(&tc->t_tcp_node);
-       spin_unlock_bh(&rds_tcp_conn_lock);
+       spin_unlock_irqrestore(&rds_tcp_conn_lock, flags);
 
        kmem_cache_free(rds_tcp_conn_slab, tc);
 }
@@ -308,13 +309,13 @@ static int rds_tcp_conn_alloc(struct rds_connection *conn, gfp_t gfp)
                rdsdebug("rds_conn_path [%d] tc %p\n", i,
                         conn->c_path[i].cp_transport_data);
        }
-       spin_lock_bh(&rds_tcp_conn_lock);
+       spin_lock_irq(&rds_tcp_conn_lock);
        for (i = 0; i < RDS_MPATH_WORKERS; i++) {
                tc = conn->c_path[i].cp_transport_data;
                tc->t_tcp_node_detached = false;
                list_add_tail(&tc->t_tcp_node, &rds_tcp_conn_list);
        }
-       spin_unlock_bh(&rds_tcp_conn_lock);
+       spin_unlock_irq(&rds_tcp_conn_lock);
 fail:
        if (ret) {
                for (j = 0; j < i; j++)
@@ -527,7 +528,7 @@ static void rds_tcp_kill_sock(struct net *net)
 
        rtn->rds_tcp_listen_sock = NULL;
        rds_tcp_listen_stop(lsock, &rtn->rds_tcp_accept_w);
-       spin_lock_bh(&rds_tcp_conn_lock);
+       spin_lock_irq(&rds_tcp_conn_lock);
        list_for_each_entry_safe(tc, _tc, &rds_tcp_conn_list, t_tcp_node) {
                struct net *c_net = read_pnet(&tc->t_cpath->cp_conn->c_net);
 
@@ -540,7 +541,7 @@ static void rds_tcp_kill_sock(struct net *net)
                        tc->t_tcp_node_detached = true;
                }
        }
-       spin_unlock_bh(&rds_tcp_conn_lock);
+       spin_unlock_irq(&rds_tcp_conn_lock);
        list_for_each_entry_safe(tc, _tc, &tmp_list, t_tcp_node)
                rds_conn_destroy(tc->t_cpath->cp_conn);
 }
@@ -588,7 +589,7 @@ static void rds_tcp_sysctl_reset(struct net *net)
 {
        struct rds_tcp_connection *tc, *_tc;
 
-       spin_lock_bh(&rds_tcp_conn_lock);
+       spin_lock_irq(&rds_tcp_conn_lock);
        list_for_each_entry_safe(tc, _tc, &rds_tcp_conn_list, t_tcp_node) {
                struct net *c_net = read_pnet(&tc->t_cpath->cp_conn->c_net);
 
@@ -598,7 +599,7 @@ static void rds_tcp_sysctl_reset(struct net *net)
                /* reconnect with new parameters */
                rds_conn_path_drop(tc->t_cpath, false);
        }
-       spin_unlock_bh(&rds_tcp_conn_lock);
+       spin_unlock_irq(&rds_tcp_conn_lock);
 }
 
 static int rds_tcp_skbuf_handler(struct ctl_table *ctl, int write,