net: Fix percpu counters deadlock

When we converted the protocol atomic counters such as the orphan
count and the total socket count deadlocks were introduced due to
the mismatch in BH status of the spots that used the percpu counter
operations.

Based on the diagnosis and patch by Peter Zijlstra, this patch
fixes these issues by disabling BH where we may be in process
context.

Reported-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Tested-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

4 files changed, 16 insertions, 7 deletions

diff --git a/net/ipv4/inet_connection_sock.c b/net/ipv4/inet_connection_sock.c
index c7cda1ca8e65..f26ab38680de 100644
--- a/net/ipv4/inet_connection_sock.c
+++ b/net/ipv4/inet_connection_sock.c
@@ -633,8 +633,6 @@ void inet_csk_listen_stop(struct sock *sk)
 
                 acc_req = req->dl_next;
 
-                percpu_counter_inc(sk->sk_prot->orphan_count);
-
                 local_bh_disable();
                 bh_lock_sock(child);
                 WARN_ON(sock_owned_by_user(child));
@@ -644,6 +642,8 @@ void inet_csk_listen_stop(struct sock *sk)
 
                 sock_orphan(child);
 
+                percpu_counter_inc(sk->sk_prot->orphan_count);
+
                 inet_csk_destroy_sock(child);
 
                 bh_unlock_sock(child);
diff --git a/net/ipv4/proc.c b/net/ipv4/proc.c
index 614958b7c276..eb62e58bff79 100644
--- a/net/ipv4/proc.c
+++ b/net/ipv4/proc.c
@@ -38,6 +38,7 @@
 #include <net/tcp.h>
 #include <net/udp.h>
 #include <net/udplite.h>
+#include <linux/bottom_half.h>
 #include <linux/inetdevice.h>
 #include <linux/proc_fs.h>
 #include <linux/seq_file.h>
@@ -50,13 +51,17 @@
 static int sockstat_seq_show(struct seq_file *seq, void *v)
 {
         struct net *net = seq->private;
+        int orphans, sockets;
+
+        local_bh_disable();
+        orphans = percpu_counter_sum_positive(&tcp_orphan_count),
+        sockets = percpu_counter_sum_positive(&tcp_sockets_allocated),
+        local_bh_enable();
 
         socket_seq_show(seq);
         seq_printf(seq, "TCP: inuse %d orphan %d tw %d alloc %d mem %d\n",
-                   sock_prot_inuse_get(net, &tcp_prot),
-                   (int)percpu_counter_sum_positive(&tcp_orphan_count),
-                   tcp_death_row.tw_count,
-                   (int)percpu_counter_sum_positive(&tcp_sockets_allocated),
+                   sock_prot_inuse_get(net, &tcp_prot), orphans,
+                   tcp_death_row.tw_count, sockets,
                    atomic_read(&tcp_memory_allocated));
         seq_printf(seq, "UDP: inuse %d mem %d\n",
                    sock_prot_inuse_get(net, &udp_prot),
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index 1f3d52946b3b..f28acf11fc67 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -1836,7 +1836,6 @@ adjudge_to_death:
         state = sk->sk_state;
         sock_hold(sk);
         sock_orphan(sk);
-        percpu_counter_inc(sk->sk_prot->orphan_count);
 
         /* It is the last release_sock in its life. It will remove backlog. */
         release_sock(sk);
@@ -1849,6 +1848,8 @@ adjudge_to_death:
         bh_lock_sock(sk);
         WARN_ON(sock_owned_by_user(sk));
 
+        percpu_counter_inc(sk->sk_prot->orphan_count);
+
         /* Have we already been destroyed by a softirq or backlog? */
         if (state != TCP_CLOSE && sk->sk_state == TCP_CLOSE)
                 goto out;
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index 10172487921b..9d839fa9331e 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -51,6 +51,7 @@
  */
 
 
+#include <linux/bottom_half.h>
 #include <linux/types.h>
 #include <linux/fcntl.h>
 #include <linux/module.h>
@@ -1797,7 +1798,9 @@ static int tcp_v4_init_sock(struct sock *sk)
         sk->sk_sndbuf = sysctl_tcp_wmem[1];
         sk->sk_rcvbuf = sysctl_tcp_rmem[1];
 
+        local_bh_disable();
         percpu_counter_inc(&tcp_sockets_allocated);
+        local_bh_enable();
 
         return 0;
 }